Fact-checked by Grok 2 weeks ago

Tesla Autopilot


Tesla Autopilot is a suite of advanced driver-assistance system (ADAS) features developed by Tesla, Inc., that enables semi-autonomous vehicle control, including adaptive cruise control, lane centering, automatic lane changes, and traffic-aware navigation, with the goal of improving safety and convenience while requiring constant driver supervision. Introduced via software update 7.0 in October 2015 for compatible Model S and Model X vehicles equipped with Autopilot Hardware 1.0, it has since expanded to all Tesla models and evolved through multiple hardware iterations, culminating in the AI4 computer and vision-only sensing without reliance on lidar or radar in recent versions.
Tesla reports that Autopilot engagement correlates with significantly lower crash rates, recording one crash per 6.36 million miles driven in Q3 —approximately six times safer than the U.S. national average of one crash per 1.03 million miles and over six times safer than vehicles driven without (one per 1.03 million miles). These metrics are derived from billions of miles of real-world fleet data, emphasizing empirical safety improvements through over-the-air software updates and training. Notwithstanding these safety statistics, has faced regulatory scrutiny, including multiple investigations by the (NHTSA) into crashes involving the system, such as failures to detect obstacles or violations of traffic controls, leading to recalls affecting millions of vehicles and ongoing probes into Full Self-Driving (Supervised) capabilities as of 2025. Critics highlight incidents where driver over-reliance contributed to fatalities, though maintains that misuse and external factors play causal roles in many cases, underscoring the system's design limitations as a supervised assistance tool rather than fully autonomous.

Historical Development

Initial Partnerships and Launch (2014–2016)

initiated its development through a partnership with , integrating the Israeli firm's EyeQ3 processor with radar, cameras, and ultrasonic sensors into Model S vehicles starting in September 2014. This hardware configuration, retrospectively termed Autopilot Hardware 1 (AP1), provided the foundation for advanced driver-assistance features focused on highway driving. In October 2015, released software version 7.0, activating the beta version of 1.0 for eligible Model S owners. Core features included traffic-aware , which adjusts speed based on forward vehicles, and autosteer, enabling lane-keeping on divided highways up to 90 mph under driver supervision. The system also supported driver-initiated automatic lane changes and forward collision warnings, with emphasizing its role in reducing driver fatigue during long-distance travel. The collaboration with dissolved in July 2016, following disputes over liability allocation, the pace of deployment, and Tesla's intent to leverage anonymized fleet data for training—approaches deemed premature after a May 2016 fatal crash involving . cited risks to its technology's reputation, while Tesla attributed the rift to 's resistance to independent evolution. By late 2016, had accumulated over 200 million miles of customer-driven engagement, with Tesla's preliminary analyses indicating crash rates several times lower than manual driving on highways, though investigations highlighted lapses as a factor in incidents.

Expansion and Rebranding (2016–2019)

In October 2016, transitioned to fully in-house development of its system following the termination of its partnership with earlier that year, introducing 2 across all new vehicles produced from that point onward. This hardware suite was marketed as enabling "full self-driving" capabilities through future software updates, with CEO stating that it provided the necessary computing power and sensors for complete autonomy. As part of this expansion, launched the Enhanced (EAP) package for approximately $5,000, adding features such as automatic lane changes, parallel and perpendicular parking (Autopark), and Summon for remote vehicle maneuvering. The Full Self-Driving (FSD) Capability package was simultaneously offered as an additional $3,000 upfront for hardware plus $5,000–$8,000 for promised software enabling urban navigation, and response, and interchanges, positioning it as a to accelerate development. These packages represented a strategic decision to monetize anticipated ahead of software maturity, with over-the-air () updates enabling incremental feature rollouts without changes. EAP software began deploying in December 2016, initially to early adopters, expanding Autopilot's scope from highway-centric operation to more versatile assisted driving while requiring driver supervision. This approach allowed to scale user adoption rapidly, as vehicles with compatible could receive enhancements fleet-wide, fostering a feedback loop for refinement through aggregated usage . A core element of this period's growth was the amplification of real-world from Tesla's expanding vehicle fleet, leveraging the eight-camera vision system in Hardware 2 to capture video clips of driving scenarios. By 2019, this had amassed billions of miles of anonymized data, uploaded selectively when vehicles encountered novel or edge-case situations, to train neural networks for and decision-making via end-to-end . This —wherein more deployed vehicles generated superior training datasets—differentiated from competitors reliant on simulated or limited real-world inputs, enabling iterative improvements in software capable of handling diverse environments. At the Autonomy Day event on April 22, 2019, underscored its commitment to vision-based autonomy, announcing plans for a shared network powered by FSD-equipped vehicles and highlighting the fleet's advantage as key to surpassing human driving performance. Amid heightened regulatory inquiries from the (NHTSA) into Autopilot's deployment and marketing, emphasized ongoing software validation while maintaining sales of FSD packages, reflecting a calculated risk to build scale despite timelines extending beyond initial projections. This phase solidified Autopilot's evolution from basic assistance to a platform poised for broader commercialization, driven by in-house and user-generated rather than external partnerships.

Recent Milestones (2020–2025)

In 2020, initiated limited releases of its Full Self-Driving (FSD) software to select owners, marking an early expansion of advanced driver-assistance capabilities beyond basic features, with initial versions focusing on urban driving scenarios under supervision. By 2021, announced a transition to a vision-only approach for and FSD, eliminating reliance on sensors in new vehicles to streamline through camera-based neural networks, a shift implemented progressively across the fleet. This period also saw the gradual broadening of FSD access, with software iterations improving handling of complex maneuvers like unprotected left turns. From 2022 to 2023, expanded FSD Beta to a wider North American user base, culminating in the November 2022 wide release of version 10.69, which enabled subscription access for qualifying owners and emphasized supervised operation in diverse environments. Version 11, rolled out broadly in early 2023, refined path planning and intervention prediction, setting the stage for subsequent advancements. The introduction of end-to-end s in FSD v12 later that year represented a , replacing modular with unified models trained on vast video datasets to directly map perceptions to vehicle controls, enhancing behavioral realism in city streets and highways. In 2024 and 2025, FSD software progressed through versions 13 and 14, incorporating larger parameter models—scaling to hundreds of billions—for improved and smoother trajectories, with v14 emphasizing supervised to mitigate edge cases while awaiting regulatory unsupervised deployment. reopened one-time FSD transfers from existing to new on April 24, 2025, applying to fully paid FSD purchases to incentivize upgrades amid ongoing software maturation. The company's Q3 2025 Safety Report documented one crash per 6.36 million miles driven with engaged, compared to higher crash rates without it or in the U.S. average, attributing gains to iterative software refinements despite increased feature complexity.

Hardware Iterations

Hardware 1 (AP1 with )

Tesla's first-generation Autopilot hardware, designated as Hardware 1 or AP1, equipped Model S and Model X vehicles produced from September 2014 to October 2016. This system integrated a single forward-facing camera for visual processing, a first-generation forward unit with a detection range of approximately 525 feet, and 12 ultrasonic sensors each capable of detecting obstacles up to 16 feet away. The sensor suite relied on these components to enable basic driver-assistance functions, without the multi-camera arrays or rear/side vision found in subsequent iterations. The core computing was handled by Mobileye's EyeQ3 processor, which performed real-time image recognition and fusion of radar data to support and lane-keeping assistance, known as Autosteer. This setup processed inputs primarily for straight-line driving, where clear markings and consistent allowed reliable within speed limits up to 90 on compatible roads. The EyeQ3's architecture emphasized cost-effective vision-based detection but lacked redundancy for diverse environmental inputs, constraining its deployment to controlled, divided- environments. AP1 exhibited limitations in non-highway scenarios, such as intersections or roads with faded markings, due to its forward-only and absence of high-resolution side or rear detection, which could lead to incomplete in curves, merges, or low-visibility conditions like or . In response to a fatal in May involving a Model S on , where the system failed to detect a crossing tractor-trailer against a bright , the initiated an investigation, prompting to issue a voluntary software affecting approximately 29,000 vehicles in July ; this enhanced driver engagement cues and logging without altering the hardware. Early safety reports indicated empirical reliability on s, with Autopilot-engaged miles showing crash rates as low as one incident per 5.3 million miles in late , outperforming the U.S. average of one per 94,000 miles without the system, though these figures were derived from user-reported fleet limited to highway use.

Hardware 2 and 2.5 (AP2)

introduced Hardware 2 (also known as AP2) in October 2016 for new Model S and Model X vehicles, following its split from earlier that year. This represented a shift to -designed hardware, incorporating the NVIDIA DRIVE PX 2 platform customized for greater computational capacity to support processing for autonomous driving. The system featured eight cameras providing 360-degree visibility, a forward-facing , and twelve ultrasonic sensors, expanding sensor redundancy beyond the single forward camera in Hardware 1. The AP2 compute module delivered substantially higher processing power than its predecessor, facilitating the and of vision-based neural networks essential for advanced perception tasks. This upgrade enabled Tesla to pursue full self-driving capabilities independently, with the hardware's architecture optimized for handling large-scale data from the expanded camera array and radar fusion. In August 2017, Tesla rolled out Hardware 2.5 (AP2.5) as an incremental upgrade, adding a secondary compute node to enhance overall processing power and introduce redundancy in both computation and wiring harnesses. This dual-processor setup improved , addressing potential single points of failure in the original AP2 design while maintaining compatibility with the existing sensor suite. AP2.5 vehicles continued to rely on radar-camera fusion for and path planning, though the architecture's emphasis on compute laid groundwork for iterative over-the-air enhancements.

Hardware 3 (FSD Computer)

introduced Hardware 3 (HW3), also known as the Full Self-Driving (FSD) Computer, in April 2019 during its Day event, marking a shift to in-house developed for advanced driver assistance and potential full . The system features two custom-designed neural processing units (NPUs), each capable of 72 tera operations per second (), delivering a combined 144 for inference tasks such as and path planning. This compute power enables processing of up to 2,300 camera frames per second, supporting 's vision-based perception approach. HW3 incorporates through dual system-on-chips (SoCs), allowing seamless if one unit fails, along with redundant supplies to enhance reliability for safety-critical operations. Each includes 12 CPUs, a GPU for visualization, and 8 GB of LPDDR4 , optimized for low consumption at around 72 watts for the . began in late 2019, with installation standard in all new vehicles from that period onward, equipping millions of cars produced between 2019 and early 2023 before the transition to HW4. Tesla positioned HW3 as sufficient for achieving autonomy, including operations, with CEO stating it would enable full self-driving capabilities without needing further hardware upgrades. HW3-equipped vehicles contributed to fleet-wide , processing sensor inputs in shadow mode to generate training datasets for improvements in early FSD releases. By 2025, however, questions arose regarding HW3's adequacy for driving, prompting lawsuits from owners who purchased FSD packages expecting robotaxi-level performance. In and , class actions alleged misleading claims about HW3's capabilities, leading to concede that retrofits to newer hardware might be necessary for FSD buyers if proves unattainable on HW3. has indicated potential free upgrades for affected FSD purchasers post-validation of superior hardware, amid ongoing debates over the original promises' verifiability.

Hardware 4 (AI4)

Tesla's Hardware 4 (HW4), also referred to as AI4, represents an incremental advancement over Hardware 3, emphasizing enhanced resolution and computational to support advanced driver-assistance features. Production vehicles equipped with HW4 began shipping in 2023, initially integrated into refreshed Model S and Model X sedans and SUVs starting in February of that year. The system maintains Tesla's vision-only perception approach but incorporates upgraded cameras with resolutions up to 5 megapixels—compared to 1.2 megapixels in HW3—enabling crisper imagery for distant and finer detail extraction. Specific camera specs include a front-facing unit at 2896 x 1876 pixels and a rear camera at 1448 x 938 pixels, which facilitate improved low-light performance and through reduced noise and higher . HW4's Full Self-Driving computer delivers approximately three to four times the processing power of HW3, with a peak power draw of around 160 watts during intensive operations, allowing for faster inference on models without relying on external hardware upgrades. This boost in compute supports in and fault-tolerant processing, mitigating risks in failure scenarios by distributing workload across dual nodes similar to HW3 but with greater headroom for future software iterations. Deployment expanded to the Cybertruck upon its production start in November 2023, as well as refreshed Model Y variants and the updated Model 3 (), ensuring all new vehicles from mid-2023 onward feature HW4 as standard. In vehicle testing and user-reported data, HW4 configurations have demonstrated lower disengagement rates in Full Self-Driving supervised mode, with community trackers logging averages exceeding 300 miles per critical intervention in urban environments on software versions like v13, attributed to the 's superior over HW3 equivalents. These improvements stem from the 's ability to process higher-fidelity inputs, though has not released official disengagement statistics segmented by hardware version, relying instead on aggregate reports that show Autopilot-enabled vehicles achieving one crash per millions of miles driven. Critics note that while HW4 provides marginal redundancy gains, its vision-centric design lacks diverse sensor backups like , potentially limiting robustness in adverse .

Hardware 5 (AI5) and Future Prospects

Tesla introduced Hardware 5, rebranded as AI5, as its next-generation Full Self-Driving computer, with announcing in September 2025 that the chip delivers up to 8 times the raw compute power of the AI4 predecessor, alongside 9 times more memory and enhancements up to 40 times in select performance metrics. Leaked specifications indicate AI5 achieves 2000–2500 ( operations per second), representing roughly 5 times the inference performance of AI4 while running operations 10 times more cost-effectively than comparable chips. Manufacturing will occur at facilities in and plants in , enabling scaled production for integration into vehicles like the Cybercab starting in early 2026. AI5 incorporates hardware optimizations tailored for Tesla's vision-based autonomy stack, including upgraded camera sensors and dedicated systems for front-camera maintenance to sustain clear visibility in adverse conditions. These features address empirical challenges in long-duration operations, such as accumulation of road grime or precipitation on forward-facing lenses, through automated wiper-spray sequences that precisely target the camera enclosures without relying on manual intervention. Production ramp-up prioritizes robotaxi fleets, where uninterrupted sensor fidelity directly correlates with operational reliability, as validated in Tesla's internal testing of similar cleaning mechanisms. Looking ahead, AI5 positions to accelerate validation of Full Self-Driving on broader fleets, building on AI4's capabilities once core autonomy milestones are met. has projected that the compute scaling in AI5 will yield substantial safety improvements by enabling more sophisticated , potentially reducing disengagement rates through refined end-to-end learning models trained on expanded datasets. These gains stem from first-principles scaling laws in AI, where increased (floating-point operations per second) empirically correlate with higher model accuracy in and tasks, as observed in 's iterative deployments. However, realization depends on software and regulatory hurdles, with emphasizing oversupply of AI5 units to support both vehicular and data-center redundancy.

Software Packages and Features

Basic Autopilot Capabilities

Basic Autopilot provides two primary (ADAS) features standard on vehicles: Traffic-Aware Cruise Control (TACC) and Autosteer. TACC, introduced in January 2015, automatically adjusts the vehicle's speed to maintain a driver-selected following distance from the vehicle ahead, or to a set speed if no leading vehicle is detected, enhancing driving by reducing acceleration and braking inputs. Autosteer, rolled out as part of the initial suite in October 2015, uses cameras and sensors to detect lane markings and keep the vehicle centered within its lane on multi-lane divided s with clear markings, requiring driver hands on the wheel and periodic torque application to confirm attentiveness. These features operate primarily on well-marked highways and do not include city streets or complex maneuvers. Since April 2019, all new vehicles have shipped with Basic enabled, allowing activation via the right scroll wheel on the steering yoke or wheel. delivers incremental improvements to Basic through over-the-air () software updates, which have reduced issues like phantom braking and improved lane-keeping smoothness over time. User reports and studies indicate Basic Autopilot correlates with reduced driver mental and physical strain during extended highway drives. A 2021 survey found users experienced less compared to , attributing this to decreased workload from sustained speed and lane maintenance. However, drivers must remain vigilant, as the system disengages if no input is detected for prolonged periods.

Enhanced Autopilot (EAP)

Enhanced Autopilot (EAP) is an optional software package offered by that extends basic capabilities with advanced driver assistance features focused on highway navigation and low-speed maneuvers, requiring constant driver supervision. Introduced in late 2016 for vehicles equipped with 2, EAP includes functionalities such as automatic lane changes, Navigate on Autopilot, Autopark, and Summon, distinguishing it from standard 's traffic-aware and lane-keeping by adding route-based decision-making and parking automation. These features aim to reduce driver workload during long highway drives and in parking scenarios, though they do not enable unsupervised operation or urban street handling. Navigate on , a core EAP feature, was first released on October 24, 2018, via software version 9.0 (2018.42), enabling the vehicle to suggest and execute lane changes to follow routes, pass slower vehicles, and take exits or interchanges while providing visual and audible alerts to the driver. An update on April 3, 2019, made it "more seamless" by reducing the need for driver confirmations in certain scenarios and expanding availability to a wider fleet, after which drivers had logged over 66 million miles with the feature. Subsequent software iterations through 2025 have refined its performance, including smoother trajectory planning and integration with high-definition maps for better route adherence, though it remains limited to pre-mapped . For low-speed conveniences, EAP incorporates Autopark, which detects and maneuvers into or spaces using ultrasonic sensors and cameras, and Summon modes including Dumb Summon for straight-line forward or reverse movement up to 12 meters via the mobile app. Smart Summon, added later, allows the vehicle to navigate obstacles in parking lots to reach the owner, initially requiring line-of-sight but evolving through software updates to handle more complex paths while the driver supervises via the app. These features, available since around 2018, have seen incremental improvements in and obstacle detection via over-the-air updates, enhancing reliability in varied environments without shifting to full . As of 2025, EAP is priced at $6,000 as a one-time purchase , positioned between free Basic and the $12,000 Full Self-Driving package, with subscription upgrades to higher tiers available for $99 per month. Adoption data specific to EAP is limited, but 's overall advanced driver assistance take rates hover around 20-30% for paid options, reflecting selective uptake due to the package's intermediate scope and the company's shifting focus toward Full Self-Driving. periodically adjusts availability, briefly discontinuing EAP in April 2024 before reintroducing it, underscoring its role as a bridge for users seeking enhanced highway and parking aids without committing to city-driving capabilities.

Full Self-Driving (FSD) Suite

The Full Self-Driving (FSD) suite, introduced by in October 2016, represents the company's premium autonomy package designed to enable complete vehicle operation without human intervention, targeting autonomy across diverse environments. Unlike Enhanced , which primarily augments highway navigation with features such as automatic lane changes and Navigate on Autopilot, FSD extends capabilities to unstructured urban settings, including the ability to handle traffic signals, stop signs, and complex intersections. This announcement coincided with equipping all new vehicles with dedicated hardware, including upgraded cameras and computing, to support eventual unsupervised operation. Core FSD functionalities emphasize city-driving proficiency, such as autosteering on residential and urban roads, responsive navigation around pedestrians and cyclists, and dynamic route adjustments for obstacles like construction zones. The system processes visual inputs to execute maneuvers including unprotected left turns, yielding to vehicles, and in varied conditions, distinguishing it from highway-centric aids by addressing the higher variability of non-freeway scenarios. These ambitions position FSD as a foundational step toward applications, though realization has depended on iterative software refinements. As of , FSD operates exclusively in supervised mode, mandating constant driver attention via cabin monitoring and torque requirements to ensure hands-on readiness, despite marketing as a pathway to full . This persistent supervision reflects ongoing limitations in edge-case handling and regulatory hurdles, with features like hands-free engagement indicators introduced to enhance but not eliminate oversight. Adoption has faced headwinds, evidenced by year-over-year declines in FSD-related in Q3 2025, attributed to lapping prior one-time boosts from vehicle releases rather than core sales growth, amid broader profitability pressures from investments in infrastructure.

Pricing, Subscriptions, and Transfers

has offered Full Self-Driving (FSD) capability via one-time purchases or monthly subscriptions, with prices adjusted periodically to reflect development progress and market demand. The one-time FSD purchase price historically ranged from $10,000 to $15,000 before dropping to $12,000 in 2023 and further to $8,000 in April 2024, where it stabilized through 2025. Subscription pricing followed suit, reducing from $199 to $99 per month in April 2024 to broaden accessibility, remaining at that level into 2025. FSD transfers between vehicles, typically non-transferable upon sale or trade-in, have been enabled temporarily as sales incentives. In April 2025, Tesla reopened transfers for FSD (Supervised), permitting owners to move the license from an eligible current vehicle to a new one delivered on or after April 24, 2025, provided the source vehicle is traded in and FSD was purchased outright. This limited-time program, without a fixed end date initially but concluding by September 30, 2025, tied into broader purchase incentives like 0% APR financing on select models to boost deliveries amid maturing software. FSD revenue trends reflect these dynamics, with Q3 2025 showing a year-over-year decline in one-time FSD recognition after lapping elevated prior-period , even as overall rose 12% to $28.1 billion. This drop coincided with FSD maturation, reducing upfront uptake as subscribers awaited enhancements linked to potential regulatory approvals for expanded , thereby influencing the package's perceived long-term value. Services and other , including software, grew 25% to $3.5 billion, but FSD's episodic pricing pressures highlighted reliance on subscriptions for recurring streams.

Technical Approach

Vision-Only Perception System

's vision-only perception system, branded as Tesla Vision, adopts a camera-centric designed to replicate human-like visual processing, prioritizing software-derived redundancy over hardware sensors such as or . This approach posits that cameras, augmented by neural networks, can achieve superior generalization and scalability compared to multi-sensor fusion, which executives have described as a costly "" that fails to address fundamental perception challenges like or dynamic environments. By relying solely on visual input, the system aims for end-to-end learning that infers depth, velocity, and semantics directly from image streams, avoiding the calibration complexities and data silos inherent in fusing disparate sensor modalities. The transition to pure Tesla Vision accelerated in May 2021, when ceased equipping new Model 3 and Model Y vehicles with forward-facing , extending the removal to Model S and X by early 2022; this shift was motivated by empirical observations from fleet data indicating that introduced fusion errors and false positives, particularly in scenarios involving curved roads or clutter. Vehicles employ eight exterior cameras—three forward-facing with varying focal lengths for wide, main, and narrow fields of view, plus side repeater, rear, and pillar cameras—to deliver 360-degree coverage up to approximately 250 meters. is computed via estimation networks that analyze sequential frames for , disparity cues, and learned priors, enabling 3D occupancy mapping without stereoscopic hardware or direct ranging. Fleet-wide deployment has provided empirical validation, with vision-only updates demonstrating reduced false braking in adverse conditions; for instance, post-radar removal software iterations exhibited greater confidence in , maintaining highway speeds without the excessive decelerations seen in sensor-fused predecessors. In fog or , where point clouds degrade due to (reducing by up to 25-50% per controlled studies), Tesla's neural networks leverage contextual cues like texture gradients and motion , corroborated by over 10 billion real-world miles of showing parity or superiority in disengagement rates for weather-impacted drives. Critics arguing the omission of lidar undermines safety—citing its precise ranging in clear conditions—are countered by Tesla's accumulation of equivalent simulated miles, exceeding 100 billion annually via physics-based rendering that replicates rare edge cases (e.g., sudden crossings in low visibility) far beyond what physical testing could achieve cost-effectively. This simulation equivalence, validated against real-world interventions, underscores the philosophy that vision scales with compute and data volume, rendering 's incremental benefits marginal once networks internalize human-equivalent invariances.

Neural Network Training and End-to-End Learning

Tesla's architecture for and Full Self-Driving evolved from modular systems, which separated , , and into distinct components reliant on hand-coded rules, to end-to-end learning paradigms that raw inputs—primarily camera feeds—directly into vehicle control outputs such as , , and braking. This transition, implemented in Full Self-Driving version 12 released in 2024, eliminated approximately 300,000 lines of explicit C++ code in favor of a unified trained to infer causal from vast datasets, enabling more robust handling of nuanced driving scenarios that rule-based modules often failed to anticipate due to their rigidity in edge cases. End-to-end networks prioritize learning implicit causal relationships between environmental inputs and actions, akin to drivers developing through , rather than decomposing tasks into potentially misaligned sub-modules that can introduce compounding errors or overlook interdependent factors like dynamics and intent. occurs via on video clips and from Tesla's global fleet, accumulating billions of real-world miles annually to capture diverse conditions, supplemented by simulated miles to augment without real-world risk. This data-driven approach allows to generalize beyond programmed heuristics, as evidenced by improved performance in unstructured environments where modular systems previously relied on brittle heuristics prone to in novel situations. In 2025, Full Self-Driving version 14 introduced a tenfold increase in parameters, enhancing capacity for modeling complex, low-probability scenarios and yielding projected exponential gains in reliability by refining the model's ability to predict and respond to causal chains in dynamic . This scaling, informed by iterative training on expanded sets, underscores a commitment to architectures that derive decisions from probabilistic over environmental , mitigating the limitations of earlier rule-based interventions that could not scale with the variability of real-world .

Dojo Supercomputer and Data Infrastructure

Tesla's supercomputer was designed as a custom-built system for large-scale neural networks on from its fleet, emphasizing high-efficiency tailored to tasks. Unveiled at Tesla's AI Day event on August 19, 2021, incorporates proprietary chips optimized for matrix multiplications and video decoding, with an "exa-pod" configuration of 120 tiles delivering approximately 1.1 exaFLOPs of compute performance at BF16 precision. This architecture aimed to handle petabytes of unstructured driving footage, enabling end-to-end model without reliance on general-purpose GPUs. The supporting data infrastructure centers on shadow mode, a passive testing regime deployed across Tesla's global fleet of over 5 million vehicles as of 2025, where Full Self-Driving software simulates control decisions in parallel with the human driver without intervening. This collects vast unlabeled datasets—exceeding billions of miles annually—by logging prediction errors, near-misses, and environmental variations only when discrepancies arise, minimizing upload volumes while capturing rare edge cases for iterative refinement. In-house processing via was intended to maintain data privacy by avoiding transmission of raw video to third-party clouds, reducing latency in feedback loops and costs associated with external bandwidth. Dojo's efficiency targeted faster training cycles compared to GPU clusters, with claims of up to 1.5 petaFLOPs per kilowatt in FP16, potentially accelerating and contributing to empirical safety gains through rapid incorporation of fleet-learned behaviors. However, in August 2025, disbanded the development team, including lead architect Peter Bannon, redirecting resources to commercial hardware from , , and others for AI training, including support for FSD version 14 released that October. This shift prioritizes scalability via established vendor ecosystems over custom silicon, though it increases dependence on external compute amid ongoing fleet data ingestion.

Full Self-Driving Capabilities

Supervised FSD Versions (v12–v14)

Full Self-Driving (FSD) Supervised versions 12 through 14 represent iterative advancements in 's beta software, requiring constant driver oversight and manual intervention as needed, with deployment limited to compatible hardware like HW3 and HW4 vehicles. These versions emphasize end-to-end architectures, shifting from rule-based heuristics to AI-driven control for more human-like maneuvers in urban and highway environments. Rollouts began with v12 in early 2024, progressing to wider availability in v13 and v14 by mid-2025, incorporating refinements in , merging, and speed adaptation. adjusted the driver monitoring "strike" system in update 2025.32, reducing the forgiveness window for inattentiveness alerts from seven days to 3.5 days per strike, after which five accumulated strikes suspend FSD use for one week. Version 12, starting with subversions like v12.5.6 in October 2024, implemented end-to-end learning across city streets and highways, enabling smoother acceleration, turning, and obstacle avoidance without modular coding. This update extended end-to-end processing to highway driving for all models, improving merge confidence by anticipating speed changes. Initial HW3 compatibility arrived in late 2024 with v12.6, addressing older hardware limitations while maintaining supervised operation. FSD v13, rolling out in early 2025, enhanced capacity with features like activation from a parked state, integrated unpark and reverse maneuvers, and refined speed profiles for reduced hesitation. v13.2 introduced 3x longer context processing, audio input integration for environmental cues, and better reward modeling to minimize false braking. Highway merging saw improvements in handling on-ramps with variable speeds, alongside camera cleaning optimizations. By October 2025, v14 subversions such as v14.1.4 in update 2025.32.8.16 added arrival options for precise destination , customizable speed profiles, and enhancements for better visualization. Undocumented refinements included advanced Autopark capabilities and reduced "brake stabbing" for smoother stops, with a modified "lite" variant planned for broader HW3 access in late 2025. User demonstrations reported drives exceeding 300 miles with zero interventions under supervision, particularly in suburban settings where over 90% of segments required no driver input.

Transition to Unsupervised Autonomy

Tesla's transition to unsupervised Full Self-Driving (FSD) relies on rigorous validation processes, including software-in-the-loop simulations that test edge cases derived from fleet data encompassing billions of real-world miles. These simulations enable the identification and mitigation of rare scenarios, such as unusual behaviors or adverse interactions, which occur infrequently in live but are amplified through accelerated virtual testing to ensure statistical reliability before deployment. Fleet vehicles contribute anonymized data from millions of users, allowing Tesla to observe and retrain neural networks on low-probability events, gradually reducing rates in supervised mode as empirical margins improve. By mid-2025, outlined plans to initiate FSD operations in select geofenced U.S. cities, targeting areas like parts of and where regulatory environments are more permissive, with rollout expected by year-end for hardware-capable vehicles. This phased approach prioritizes contained environments to validate disengagement-free performance empirically, building on supervised FSD versions (v12–v14) that have demonstrated progressive reductions in driver interventions through end-to-end refinements. However, hardware limitations, such as Hardware 3 (HW3) vehicles' inability to support capabilities without upgrades, have constrained broader deployment, highlighting technical constraints over regulatory ones as the binding factor. Despite optimistic timelines from leadership, empirical evidence points to technological unreadiness as the primary impediment, evidenced by ongoing NHTSA investigations into over 50 FSD-related traffic violations, including red-light incursions and improper lane usage in 2.9 million vehicles as of October 2025. Regulatory hurdles exist, particularly in and , but in U.S. states with minimal barriers, persistent edge-case failures and safety probes underscore that data-driven iteration—while advancing—has not yet achieved the requisite reliability for widespread unsupervised use, necessitating further simulation-validated improvements rather than external approvals as the causal bottleneck.

Robotaxi Deployment Plans

Tesla unveiled the Cybercab, a dedicated two-passenger lacking a or pedals, on , 2024, with production slated to begin in the second quarter of 2026 at an estimated cost under $30,000 per unit. The centers on a shared network where owners can opt-in their cars for ride-hailing when idle, supplemented by company-owned Cybercab fleets, enabling with operators through low operational costs absent human drivers. Deployment plans target an initial unsupervised pilot in , by the end of 2025, operating without safety monitors or additional human occupants, starting small and scaling to broader unsupervised operations. Users hail rides via the Tesla Robotaxi app, which allows destination input, ride confirmation, and notifications for pickup, with service availability expanding post-Austin to cities like . Scaling beyond the pilot faces delays, with full Cybercab production and widespread fleet rollout projected into amid production ramp-up and internal adjustments, potentially pushing millions of autonomous vehicles online by year-end. This timeline reflects moderated expectations from earlier ambitions, prioritizing safety validation through initial supervised phases before unsupervised expansion. The model promises economic disruption to traditional ride-hailing by slashing costs—potentially to $0.20–$0.30 per mile versus $1–$2 for human-driven services—through eliminated labor expenses and high utilization rates, enabling to capture significant market share from incumbents like . Analysts project robotaxi operations could comprise up to 90% of 's enterprise value by 2029 if utilization and pricing models succeed, though realization hinges on achieving reliable at scale.

Empirical Safety Performance

Tesla's Reported Crash Statistics (2019–2025)

Tesla's vehicle safety reports, initiated in late 2018, compile data from its global fleet via over-the-air , recording crashes as incidents involving or active restraint deployment, or disengagement within five seconds prior to impact, typically corresponding to forces from collisions at about 12 mph (20 km/h) or greater. These statistics differentiate between miles driven with engaged (including Full Self-Driving Supervised modes where applicable) and miles driven without in vehicles, excluding invalid or duplicated reports. In Q3 2025, reported one crash for every 6.36 million miles driven with engaged, compared to one crash for every 963,000 miles without . Earlier quarters in 2025 showed slightly higher figures for : 7.44 million miles per crash in Q1 and 6.69 million in Q2, with no- rates fluctuating around 1.2–1.5 million miles per crash.
QuarterAutopilot Miles per CrashNo-Autopilot Miles per Crash
Q1 20257.44 million1.51 million
Q2 20256.69 million1.26 million
Q3 20256.36 million0.96 million
Data for Full Self-Driving Supervised, a subset of usage, indicates even lower crash rates in recent reports, with Q2 2025 figures demonstrating approximately 10 times fewer crashes per mile than non- Tesla driving. Over the 2019–2025 period, miles per crash have trended upward from early figures of roughly 1–2 million in 2019 quarters to the 6–7 million range by 2025, reflecting cumulative software updates and data-driven refinements reported by . By Q3 2025, the fleet had accumulated billions of miles with engaged, enabling these self-reported metrics.

Comparisons to Human-Driven Vehicles

Tesla's vehicle safety reports, based on data from its fleet, show that in the third quarter of 2025, Autopilot-engaged driving recorded one crash (defined as deployment, , or police-reported incident) for every 6.36 million miles driven. This rate is approximately four times lower than the 1.52 million miles per crash for Tesla vehicles driven without Autopilot in the same period. Comparisons to national benchmarks further highlight the disparity: U.S. averages from NHTSA and FHWA data indicate roughly one crash per 700,000 miles across all , rendering usage about nine times safer than the baseline human-driven fleet average. These figures derive from 's aggregation of millions of miles logged quarterly, contrasted against estimates incorporating diverse conditions and types. Critics note a potential in exposure, as is primarily deployed on , where crash rates are inherently 3–5 times lower than urban or rural roads due to fewer intersections and lower speeds variability. However, even after adjusting for this highway predominance—via normalized models accounting for road-type distributions—studies on ADAS systems, including Tesla's, affirm net gains, with reducing errors in speed maintenance, lane-keeping, and collision avoidance by factors of 2–4 relative to matched . Over cumulative operation exceeding 3 billion Autopilot miles by mid-2025, Tesla's data reflect a downward trend in fleet-wide rates, attributing reductions to iterative software updates enhancing predictive braking and obstacle detection. This empirical trajectory underscores automation's strength in mitigating human factors like and , which contribute to over 90% of U.S. crashes per NHTSA analyses, though edge cases demanding human-like remain areas where supervised systems defer to drivers.

Causal Analysis of Incidents and Improvements

In analyses of Tesla Autopilot incidents, the predominant causal factor has been driver inattention or misuse, such as prolonged hands-off-wheel operation or failure to monitor the roadway, rather than inherent system defects. The (NHTSA) has documented this pattern across multiple investigations, noting that drivers often disengaged from active supervision despite Autopilot's design as a supervised assistance system requiring constant oversight. For instance, in 13 fatal crashes examined by NHTSA through 2024, misuse—including overreliance without intervention—contributed directly, with the system performing within expected parameters until driver neglect allowed escalation. System-level flaws, when present, typically manifest in rare edge cases like sun glare, , or atypical object occlusion, where camera-based perception struggles under degraded conditions. These incidents represent a minority of reported events, often compounded by driver factors, and Tesla has iteratively addressed them through over-the-air () software updates that refine predictions without hardware changes. Examples include enhancements to stationary , where early versions occasionally failed to classify unmoving vehicles or debris, leading to OTA retraining on fleet data to improve precision by up to 17% in challenging scenarios like high-curvature roads or low light. Recent advancements in Full Self-Driving (FSD) software, particularly version 14 released in 2025, have targeted persistent issues like phantom braking—unprompted deceleration due to misperceived obstacles—via end-to-end optimizations that better contextualize and inputs. These updates have demonstrably reduced such events in supervised deployments, with fleet showing smoother fault recovery and fewer interventions needed for phantom triggers. Media coverage often amplifies isolated fatalities involving , yet per-mile fatality rates for engaged usage remain below 1 per billion miles driven, substantially lower than the U.S. human average of approximately 1.3 per 100 million vehicle-miles-traveled, underscoring that , not systemic unreliability, drives the majority of outcomes. This disparity highlights how causal attribution must prioritize empirical disengagement logs and over anecdotal emphasis, with remediation enabling rapid evolution beyond initial limitations.

Criticisms and Limitations

Technical and Edge-Case Failures

Tesla's Autopilot and Full Self-Driving (FSD) systems have demonstrated vulnerabilities in perception under adverse lighting conditions, particularly sun glare directly impacting camera sensors. The National Highway Traffic Safety Administration (NHTSA) initiated a probe into approximately 2.4 million Tesla vehicles in October 2024 following reports of crashes where FSD was engaged during reduced visibility, including sun glare, with the agency noting four such incidents. User reports and independent observations corroborate these issues, with FSD frequently disengaging or issuing takeover alerts when low-angle sunlight overwhelms camera feeds, leading to temporary blindness in lane detection and object recognition. For instance, in August 2025, analyses of owner videos highlighted failures in handling low sun positions, where the system misinterprets glare as obstacles or loses lane markings entirely. In complex environments like zones, FSD versions such as v14 have exhibited hesitations in , including delayed responses to temporary , narrowed lanes, or erratic barriers. Early 2025 tester on v14.1 indicated struggles with speed adjustments and in active work areas, where the system often slows excessively or requires interventions to avoid merging conflicts. These edge cases stem from limitations in mapping updates and perception of dynamic occlusions, though Tesla has acknowledged such scenarios in , prioritizing improvements in subsequent iterations like v14.2 for better handling of and interactions. Quantitative assessments of system reliability reveal requirements varying by version and conditions, with testing in 2024 reporting approximately 75 s per 1,000 miles in FSD Beta, equating to roughly one every 13 miles—predominantly triggered by perceptual uncertainties in unstructured scenarios. While reports ongoing improvements in miles per through data-driven training, edge-case disengagements persist below levels needed for operation. Mitigations include over-the-air () software updates, enabling rapid algorithmic refinements without hardware , in contrast to competitors' frequent physical recalls for recalibrations. has deployed remedies for perception-related bugs, such as adjustments for camera handling, allowing fleet-wide fixes in days rather than months.

Driver Monitoring and Engagement Problems

Tesla's , activated via software updates starting in May 2021, utilizes the interior cabin camera—positioned above the in vehicles equipped with it since late 2020—to score driver attentiveness during or Full Self-Driving (FSD) engagement. The system detects gaze direction, head pose, and eye closure to issue escalating alerts for inattention, such as visual and auditory chimes, followed by potential strikes if ignored. Accumulating five strikes results in temporary suspension of Autopilot/FSD features, with each strike forgiven after a penalty-free period; as of software update 2025.32 in September 2025, this forgiveness interval was halved from seven days to 3.5 days per strike to encourage stricter compliance. Despite these mechanisms, driver engagement issues persist, as evidenced by U.S. (NHTSA) analyses of over 950 Autopilot-related crashes from 2018 to 2023, where the majority involved driver inattention or misuse, such as looking away from the road or failing to intervene promptly. NHTSA attributed this to the system's design encouraging overreliance, noting that early torque-based (pre-2021 camera reliance) was easily gamed with weights or braces, prompting the shift to camera-based oversight. Independent studies corroborate rising complacency, with drivers in naturalistic settings exhibiting reduced roadway glances and increased secondary tasks over time during use, heightening risks in edge cases like sudden obstacles. Empirical data from Tesla's quarterly safety reports, however, indicate that enforced engagement via monitoring correlates with lower crash rates: vehicles using Autopilot (requiring active supervision) recorded one crash per 7.63 million miles driven in Q4 2023, versus one per 1.55 million miles without Autopilot but with basic safety features, and the U.S. average of one per 670,000 miles. AAA Foundation testing found Tesla's camera system delivers consistent alert timings (37-39 seconds under varied lighting), outperforming some competitors in reliably prompting re-engagement without excessive false positives. This suggests monitoring mitigates misuse by upholding the "supervised" designation, serving as a causal safeguard rather than an inherent flaw; narratives framing human oversight as a failure overlook Level 2 automation's reliance on drivers as the ultimate failsafe, with data showing net safety gains when compliance is maintained.

Skepticism from Former Insiders and Competitors

In October , two former leaders of Tesla's self-driving and programs publicly diverged from CEO Elon Musk's optimistic assessments of progress toward unsupervised autonomy, emphasizing persistent technical hurdles and questioning the feasibility of near-term deployment despite Musk's claims of imminent breakthroughs. These comments highlight internal debates over development pace, with the ex-executives arguing that scaling reliable unsupervised operation remains elusive without fundamental architectural shifts beyond current vision-only neural networks. Peter Rawlinson, a former Tesla senior engineer involved in early Model S development who now leads , has similarly expressed skepticism about Tesla's Full Self-Driving (FSD) timelines, predicting in 2024 that unresolved core challenges in the system could persist for at least another decade, contrasting Musk's repeated assertions of solving within months or years. Rawlinson's , rooted in his direct experience with Tesla's hardware-software , underscores concerns over the limitations of Tesla's camera-reliant approach in achieving human-level across diverse environments. Competitors like have amplified these doubts, with Waymo representatives questioning the safety efficacy of Tesla's FSD in unsupervised scenarios due to reported incidents and the absence of comprehensive third-party validation comparable to Waymo's own metrics, which show disengagement rates orders of magnitude lower in operational fleets. 's leadership has highlighted their multi-sensor (lidar-inclusive) strategy as superior for robustness, implicitly critiquing Tesla's vision-only scaling as risk-prone despite Tesla's data advantage from over 6 billion miles of real-world driving logged by its vehicle fleet as of mid-2025—far exceeding Waymo's geofenced operations limited to thousands of vehicles in select cities. Such insider and rival perspectives, however, predate or undervalue Tesla's post-2023 advancements, including the v12+ transition to end-to-end learning models trained on billions of video clips, which have empirically reduced rates by factors of 5-10x in user-reported from 2024-2025, enabling smoother of unstructured scenarios that earlier modular systems struggled with. FSD take rates, hovering around 20-30% of new vehicle sales in 2025, reflect consumer caution amid high-profile scrutiny rather than validated tech deficiencies, as evidenced by surveys where 35% of respondents cited perceptions—often media-driven—as deterring factors, outweighed by the 14% attracted but indicating room for trust-building through transparent mileage-based disclosures.

NHTSA Investigations and Recalls

The (NHTSA) initiated its first formal investigation into Tesla's system in October 2016 following a fatal crash involving a Model S in , where the system reportedly failed to detect a tractor-trailer crossing . This probe expanded over time, with NHTSA documenting nearly 1,000 crashes involving or Full Self-Driving (FSD) features from January 2018 to August 2023, including 29 fatalities. By mid-2025, investigations encompassed over 1,000 reported incidents, prompting requirements for enhanced driver monitoring, such as cabin camera alerts and steering wheel torque checks to ensure hands-on engagement. These probes revealed patterns of misuse, including operation on undivided roads or in low-visibility conditions, leading to mandates for software restrictions and visual/auditory warnings. In December 2023, NHTSA ordered a affecting over 2 million vehicles (Models S, X, 3, and Y from 2016 onward) to address Autopilot's inadequate safeguards against foreseeable misuse, such as insufficient driver attentiveness checks. The remedy involved an over-the-air () software deploying recurring visual and audible alerts, reduced system availability on incompatible , and logging of inattentiveness for potential suspension of features. complied by deploying the to 2.03 million vehicles, with NHTSA later probing its effectiveness in April 2024 after post-recall crash reports, including incidents where alerts failed to prevent disengagement or collisions. Similar fixes resolved issues like low-speed maneuvers, such as unexpected sharp turns or failure to yield, by refining trajectory planning and in subsequent updates. NHTSA's scrutiny intensified in 2024–2025 amid ambitions, with probes questioning the adequacy of older Hardware 3 (HW3) systems in vehicles from 2016–2019 for unsupervised operation. A January 2025 investigation targeted 2.6 million vehicles' remote driving feature after crashes, while an October 2025 probe into 2.9 million units (covering HW3- and HW4-equipped models from 2016–2025) examined FSD's traffic violations, including 58 incidents like red-light runs and wrong-way driving, resulting in 14 crashes and 23 injuries. These actions highlighted hardware limitations in edge cases, such as degradation in legacy systems, though compliance has iteratively reduced incident rates per mile driven, as evidenced by quarterly showing Autopilot-enabled crashes at lower frequencies than national averages. Regulatory timelines have trailed rapid software iterations, yet enforced updates have demonstrably enhanced fleet-level safeguards without hardware retrofits.

Advertising and Liability Lawsuits

has faced multiple lawsuits alleging misleading advertising regarding its and Full Self-Driving (FSD) features, with plaintiffs claiming the nomenclature and marketing implied fully autonomous operation without driver intervention, akin to capability. In August 2025, a federal judge certified a lawsuit by drivers who purchased FSD, asserting that CEO Elon Musk's statements over eight years misrepresented the technology's readiness for unsupervised self-driving. Similarly, the filed suit in July 2025, accusing of deceptive practices by promoting and FSD as enabling autonomous operation despite requiring constant supervision. These claims contrast with 's explicit disclaimers since at least 2019, labeling FSD as a "beta" feature initially and later as "FSD (Supervised)," with owner manuals and software interfaces stating it demands "active driver supervision" and does not render the vehicle autonomous. Specific contention has arisen over Hardware 3 (HW3) vehicles, where previously marketed "all Tesla cars now have Full Self-Driving Hardware," but in October 2025, amid emerging lawsuits, revised this to exclude guarantees of future unsupervised capability. Plaintiffs in U.S. and cases, including a in joined by thousands of owners, argue HW3 cannot achieve promised full , breaching contracts for FSD purchases made under assurances of upgradability to Level 5. acknowledged in January 2025 that HW3 may require retrofits for unsupervised FSD, potentially obligating hardware upgrades or refunds, though outcomes have varied, with some owners securing reimbursements for unmet promises. Courts have allowed certain claims to proceed by distinguishing marketing from contractual disclaimers, yet empirical data from 's reported billions of FSD miles indicate progressive improvements in handling edge cases, supporting the view that capabilities evolve via software updates rather than absolute initial delivery. In parallel, liability lawsuits stemming from Autopilot-involved crashes have tested claims of defective or inadequate warnings, often intersecting with allegations of overstated reliability. A jury in August 2025 awarded over $240 million against in a 2019 fatal crash case, assigning 33% to for Autopilot's failure to detect and respond to hazards, despite evidence of driver inattention; rejected a $60 million prior and plans to , citing hidden showing system disengagement attempts. Other cases have resulted in settlements without admission of fault, such as undisclosed resolutions in fatal incidents, where plaintiffs emphasized insufficient driver monitoring tied to marketing that downplayed supervision needs. maintains that rests primarily with drivers for overreliance, bolstered by disclaimers and real-world safety statistics showing Autopilot-enabled vehicles with one crash per millions of miles versus higher human benchmarks, though juries have occasionally imposed partial corporate responsibility based on perceived flaws in hazard avoidance. Overall, while some suits have yielded significant verdicts or certifications, trends reflect challenges in proving causation beyond driver error, with 's iterative updates addressing identified limitations empirically rather than through absolute pre-release guarantees.

Regional Regulations and Barriers to Deployment

In the United States, (FMVSS) permit Tesla's Full Self-Driving (FSD) as a supervised Level 2 , requiring constant driver oversight, without necessitating exemptions for such features. Unsupervised deployment, however, encounters barriers including geofencing to compliant operational domains and state-specific approvals for commercial services, with the (NHTSA) providing FMVSS exemptions for automated vehicles to enable limited testing and deployment as of April 2025. Tesla has targeted unsupervised FSD rollout in 8-10 metropolitan areas by late 2025, contingent on regulatory clearances in states like , , and . European Union and regulations impose more rigorous type approval requirements under United Nations Economic Commission for Europe (UNECE) frameworks, which prioritize comprehensive prior to widespread use, delaying FSD supervised features until potentially 2028 according to Tesla's projections amid ongoing revisions. While hands-off highway operation aligns with emerging legal allowances, implementation remains stalled by precautionary validation processes that exceed those in the , fostering what Tesla executives describe as a widening regulatory disparity potentially rooted in risk-averse oversight rather than equivalent safety thresholds. In , the Ministry of Industry and (MIIT) facilitates accelerated approvals through localized data and software adaptations, positioning for supervised FSD introduction in 2025 following a temporary halt to a March trial for compliance adjustments. Such fragmented regulatory landscapes constrain Tesla's fleet-scale data accumulation, as supervised miles logged per vehicle—critical for refining neural networks against rare scenarios—remain curtailed in restrictive jurisdictions, thereby slowing empirical safety advancements that correlate with exposure volume in permissive regions.

Availability and Impact

Variations by Region (US, China, Europe)

In the United States, Full Self-Driving (Supervised) has been widely available to eligible vehicles since early 2025, incorporating features such as automated lane changes, city street navigation, and parking under driver supervision, with software updates like version 2025.32.8.16 enabling broader deployment across hardware generations including HW3 and HW4. This allows U.S. owners to access the most advanced capabilities, including end-to-end processing for complex maneuvers, subject to ongoing supervision requirements. In , Tesla has adapted Full Self-Driving through integration with 's high-definition mapping and navigation data, enabling improved localization and compliance with local data regulations as of March 2025. This partnership facilitates features like precise urban driving and highway autonomy tailored to Chinese road conditions, with engineers assisting in map data fusion to accelerate FSD version 13 and later rollouts, despite initial limitations from data export restrictions. In , basic —encompassing and lane-keeping—remains the dominant configuration, with Enhanced Autopilot and Full Self-Driving features restricted or unavailable in most markets as of October 2025 due to pending type approvals and system . Advanced capabilities like city street autosteer or hands-off highway driving are limited to supervised trials or not yet deployed, reflecting adaptations for denser urban environments and varying national standards, though UN Regulation 171 enables potential highway expansions post-September 2025. In Australia and New Zealand, Full Self-Driving (Supervised) began a phased public rollout in late August to September 2025, initially available only to vehicles equipped with Hardware 4 (HW4), such as Model 3 and Model Y, with Hardware 3 (HW3) support planned but not yet enabled. Tesla's quarterly vehicle safety reports, aggregating global data, show one crash per 6.36 million miles driven with engaged in Q3 2025, compared to one per 1.52 million miles without, with regional variations inferred from feature access: higher and FSD usage in the U.S. contributes to more miles logged and potentially lower normalized rates there, while limited advanced features in result in lower overall engagement and sparser region-specific incident data. ’s adaptations via local mapping may similarly influence safety metrics through enhanced environmental perception, though comprehensive regional breakdowns remain undisclosed in public reports.

Adoption Rates and Economic Effects

As of the third quarter of 2025, approximately 12% of Tesla's eligible vehicle fleet has activated (FSD) capability, either through outright purchase or subscription, according to company disclosures during earnings discussions. Adoption rates vary by model, reaching 50-60% among newer and owners who opt for the feature at purchase, driven by its integration with premium hardware and demonstrated utility in highway and urban scenarios. This uptake reflects Tesla's subscription model, priced at $99 monthly since 2021, which lowers barriers compared to the $8,000 one-time fee, enabling broader experimentation amid ongoing software improvements. Tesla's aggregate FSD-driven miles exceeded 3 billion cumulatively by January 2025, with quarterly FSD network mileage surging 27% quarter-over-quarter to contribute toward 1.3 billion miles in the third quarter alone, underscoring the scale of real-world that fuels iterative enhancements. Despite empirical metrics—such as one per 6.36 million Autopilot-engaged miles in Q3 2025, versus 1.71 million without—adoption remains constrained, potentially due to public skepticism amplified by selective media reporting that overlooks fleet-level statistical advantages over human benchmarks. Economically, and FSD enable cost reductions for users via Tesla Insurance, which offers up to 10% premium discounts for drivers engaging FSD Supervised at least 50% of the time, predicated on telematics-verified lower incident rates from autonomous assistance. This reflects causal links between data-driven safety gains and actuarial savings, bypassing traditional insurer biases toward conservative risk models. On a macroeconomic scale, widespread deployment could enhance by reallocating driver time, with robotaxi projections estimating a global of $43.76 billion by 2030, potentially capturing trillions in value through efficient ride-hailing and —benefits accruing from Tesla's decentralized, owner-opted fleet over centralized regulatory hurdles. Such dynamics highlight free-market mechanisms accelerating adoption beyond subsidized alternatives, fostering GDP uplift via reduced transportation frictions and novel service ecosystems.

References

  1. [1]
    Autopilot | Tesla Support
    For Tesla drivers who were not using Autopilot technology, we recorded one crash for every 963,000 miles driven. By comparison, the most recent data available ...
  2. [2]
    Autopilot - Tesla
    Autopilot · About Autopilot · Autopilot Features · Traffic Light and Stop Sign Control · Full Self-Driving (Supervised) · Autopark · Summon · Limitations and Warnings.About Autopilot · Full Self-Driving (Supervised) · Traffic Light and Stop Sign...
  3. [3]
  4. [4]
    Tesla Vehicle Safety Report
    In the 1st quarter, we recorded one crash for every 7.44 million miles driven in which drivers were using Autopilot technology. For drivers who were not using ...Tesla Canada · Tesla United Kingdom · Tesla Australia · Tesla Hong Kong
  5. [5]
  6. [6]
    Full Self-Driving (Supervised) - Tesla
    Tesla uses billions of miles of anonymous real-world driving data to train Full Self-Driving (Supervised) to take care of the most stressful parts of daily ...Menu · Tesla's FSD · Tesla Australia · Tesla Hong KongMissing: history | Show results with:history
  7. [7]
    [PDF] Additional Information Regarding EA22002 Investigation - nhtsa
    Apr 25, 2024 · The drivers were involved in crashes while using Autopilot despite fulfilling Tesla's pre-recall driver engagement monitoring criteria ...Missing: controversies | Show results with:controversies
  8. [8]
    US probes driver assistance software in 2.9 million Tesla vehicles ...
    Oct 9, 2025 · US probes driver assistance software in 2.9 million Tesla vehicles over traffic violations · NHTSA says it has 58 reports, including 14 crashes ...Missing: controversies | Show results with:controversies
  9. [9]
    The Ultimate Guide to Tesla Autopilot | Current Automotive
    Aug 31, 2020 · Autopilot Active Safety Features are included on all Tesla vehicles built since September 2014. Every vehicle with any version of Autopilot ...
  10. [10]
    Tesla recreated Autopilot's version of "Mobileye" in 6 months
    May 3, 2017 · In September 2016, Tesla and Mobileye dissolved their partnership and supply contracts and left Tesla to develop their own driving-assist ...
  11. [11]
    Tesla rolls out autopilot technology - CNBC
    Oct 14, 2015 · Tesla rolls out autopilot technology · Autosteer, which keeps Tesla vehicles in their lanes and manages their speed; · Auto lane change, which ...
  12. [12]
    Tesla reveals all the details of its 'Autopilot' and its software v7.0 ...
    Oct 14, 2015 · As previously revealed, the update includes: Automatic Emergency Steering and Side Collision Warning, 'Autosteer' (Beta), Auto lane changing and ...<|separator|>
  13. [13]
    Tesla Starts Rolling Out Autopilot Features - Bloomberg.com
    Oct 14, 2015 · Autopilot is a step toward the vision of autonomous or self-driving cars, and includes features like automatic lane changing, auto steering and ...Missing: launch | Show results with:launch
  14. [14]
  15. [15]
    Tesla says jealousy, not Autopilot safety concerns, caused breakup ...
    Sep 15, 2016 · When Mobileye found out, it “attempted to force Tesla to discontinue this development, pay them more, and use their products in future hardware, ...
  16. [16]
    How safe is Tesla Autopilot? Parsing the statistics (as suggested by ...
    Oct 13, 2016 · Counting the Chinese crash, and using Musk's latest figure for Autopilot miles driven, the Autopilot fatality rate is now one per 111 million ...
  17. [17]
    Tesla announces all production cars now have fully self-driving ...
    Oct 19, 2016 · The company announced tonight its second generation autonomous driving hardware and confirmed that it is in all vehicle off the assembly line ...
  18. [18]
    All Tesla vehicles being produced now have full self-driving hardware
    including the Model 3 — will have hardware needed for full self-driving ...Missing: 2 | Show results with:2
  19. [19]
    Tesla Enhanced Autopilot to be released in 'about 3 weeks' with ...
    Nov 27, 2016 · Tesla Enhanced Autopilot to be released in 'about 3 weeks' with incremental monthly releases | Electrek.
  20. [20]
    Tesla's upgraded Autopilot will start rolling out mid-December
    Nov 27, 2016 · Tesla's upgraded Autopilot will start rolling out mid-December. Just don't expect to get everything you want right away. Jon Fingas.
  21. [21]
    Tesla's Autopilot Depends on a Deluge of Data - IEEE Spectrum
    Aug 4, 2022 · Tesla's use of fleet vehicle data to develop Autopilot echoes the user-fueled rise of Internet giants like Google, YouTube, and Facebook. The ...
  22. [22]
    Watch how Tesla trains its neural networks for self-driving in 10 ...
    Nov 11, 2019 · Karpathy gave a quick but very informative talk about how Tesla uses machine learning to train its neural networks toward the goal of full self-driving.<|control11|><|separator|>
  23. [23]
    Elon Musk says Tesla robotaxis will hit the market next year - CNBC
    Apr 22, 2019 · All Tesla cars being produced today have the hardware on board that's required for full self-driving, Musk said, promising that, "all you need ...
  24. [24]
    Tesla unveils its new Full Self-Driving computer in detail - Electrek
    Apr 22, 2019 · Tesla unveiled all the details about its new Full Self-Driving computer, which CEO Elon Musk claims is 'objectively the best chip in the world'.
  25. [25]
    Tesla To Host Autonomy Investor Day
    Apr 3, 2019 · Tesla will host investors on the morning of April 19th at our headquarters in Palo Alto to provide a deep dive into our self-driving technology and road map.
  26. [26]
    Tesla FSD News, Software Updates, Release Notes and Statistics
    The latest on Tesla's Full Self-Driving, including software updates, news and more.FSD News · 2024.45.32.20 (FSD 12.6.4... · Additions in 2025.14.6 · 2025.2.8
  27. [27]
    Tesla FSD Beta V11 wide release gets update from Elon Musk
    Mar 14, 2023 · Tesla owners who purchased the Full Self-Driving suite would only have to wait a few more days before they receive FSD Beta V11.<|separator|>
  28. [28]
    FSD Beta goes wide release in North America, Tesla owners rush to ...
    Nov 26, 2022 · The recently released Tesla FSD Beta 10.69.3.1 (2022.36.20) is the version that Tesla owners in the United States and Canada are able to ...
  29. [29]
    Tesla pushes end-to-end neural networks for highway driving, but ...
    Nov 14, 2024 · With the latest release (v12.5.6.3), FSD is using end-to-end neural networks for driving across highways, city streets and parking lots, and ...
  30. [30]
    Tesla releases FSD v14, first major update in a year, here's what it ...
    Oct 7, 2025 · In late 2024, Tesla began rolling out FSD v13 to owners with the latest HW4 computers installed in their vehicles. ... Tesla FSD v14 release notes.Missing: v12 | Show results with:v12
  31. [31]
    Full Self-Driving (Supervised) Transfers | Tesla Support
    You can transfer an FSD (Supervised) from your current Tesla vehicle to a new Tesla vehicle for a limited time, starting April 24, 2025.
  32. [32]
  33. [33]
  34. [34]
    Tesla Autopilot AP1 vs AP2 vs AP3 - Differences Explained
    The original Tesla Autopilot system with semi self-driving capabilities was released in October of 2014 and available on the Model S, and later on the Model X.
  35. [35]
    Autopilot, Processors and Hardware – MCU & HW Demystified
    Hardware Specifications for new cars (no retrofits) ; Sonar sensors, 12, each with 16 ft range, 12 - each with 26 ft range ; Core Processors, Mobileye EyeQ3, 1 – ...
  36. [36]
    Technical hardware differences with Tesla
    Tesla bought in AP1 technology from Mobileye and it is based on their EyeQ3 hardware which is also used by a number of other makes of car for active cruise ...
  37. [37]
    Tesla Autopilot – The Ultimate Guide | Find My Electric
    Let's dive right into our Ultimate Guide to Tesla Autopilot! The Complete History of Tesla Autopilot. Truly understanding Tesla Autopilot (even as a used Tesla ...
  38. [38]
    Tesla touts Autopilot safety in first quarterly report - Smart Cities Dive
    its semi-autonomous system — experience a lower rate of accidents ...<|control11|><|separator|>
  39. [39]
    Tesla Hardware 3 (Full Self-Driving Computer) Detailed
    Tesla released its own AI self-driving hardware in 2019, dubbed Hardware 3.0 (or AP3) to power its Full Self-Driving (FSD) technology.
  40. [40]
    Tesla has a new Autopilot '2.5' hardware suite with more computing ...
    Aug 9, 2017 · Tesla's new Autopilot 2.5 suite is using a secondary node to enable more computing power, which brings the capacity closer to chipmaker's own idea of the power ...
  41. [41]
    Tesla Autonomy Day Recap – Tesla FSD - EVBite
    Apr 25, 2019 · Tesla's two-chip FSD computer at 144 TOPs would compare against the NVIDIA DRIVE AGX Pegasus computer which runs at 320 TOPS for AI perception, ...
  42. [42]
    First picture of Tesla's new Hardware 3 self-driving computer in the ...
    Jul 31, 2019 · According to the company, its new computer can process 2,300 frames per second and perform 144 trillions of operation per second. CEO Elon Musk ...<|separator|>
  43. [43]
    FSD Chip - Tesla - WikiChip
    Apr 22, 2019 · Running the full software stack, the FSD computer dissipates 72 Watts. This is roughly 25% more than the 57 Watts the prior solution, HW2.5, ...History · Overview · Full self-driving computer (FSD... · Power
  44. [44]
  45. [45]
    Elon Musk admits that Tesla will have to replace old computers for ...
    For years, Musk claimed that Tesla cars with HW3 hardware are future-proof, meaning they'll be capable of running FSD once it truly becomes what Musk wants ...
  46. [46]
  47. [47]
    Thousands of Tesla owners join class action lawsuit over 'Full Self ...
    Oct 13, 2025 · In February, shortly after Musk admitted that HW3 won't support autonomous driving, a class action was launched in Australia related to the ...Missing: sufficiency | Show results with:sufficiency
  48. [48]
    Tesla is being sued in China over not delivering self-driving to HW3 ...
    Owners in China are suing Tesla over not delivering the promised self-driving capability on vehicles equipped with Tesla's older HW3 ...
  49. [49]
  50. [50]
  51. [51]
    Tesla's new self-driving computer (HW4): more cameras, radar, and ...
    Feb 15, 2023 · New images and details surrounding Tesla's most robust self-driving computer, or Hardware 4 (HW4), were shown today, revealing additional cameras, a ...
  52. [52]
  53. [53]
    HW4 (Tesla new self-driving hardware) | Blog or Die!
    Sep 7, 2023 · Front camera resolution 2896 x 1876. Typical File size 52 MB. · Back camera resolution 1448 x 938. Typical file size 29 MB. · Side cameras ...
  54. [54]
    Tesla Hardware 4 (AI4) - Full Details and Latest News
    The new hardware, includes the following: A new front-facing camera with higher resolution and wider field of view (on some models, such as the Cybertruck).
  55. [55]
    How to Check If Your Tesla Has Hardware 3 (HW3) or Hardware 4 ...
    Aug 8, 2024 · The new cameras bring improved resolution for better details. The rear camera on AI4 has a resolution of 1448x938 vs 1280x960 on Hardware 3.
  56. [56]
    FSD HW4 v13 in a few weeks will be 25k miles per critical ... - Reddit
    Nov 22, 2024 · FSD Community Tracker -- 99% of Drives with No Critical Disengagement. 312 Miles of City Distance to Critical Disengagement, up from 126 miles.
  57. [57]
    Tesla self-driving videos don't do anything, give us the data | Electrek
    Aug 21, 2025 · The dataset currently has 40,000 miles on FSD v13.2.9, the latest FSD update on HW4 vehicles, and miles between critical disengagement currently ...
  58. [58]
    Musk Reveals Tesla AI5 Specs, Calls Next-Gen FSD Computer a ...
    Sep 11, 2025 · NEWS: Elon Musk says Tesla's AI5 chip will be up to 40X better than AI4 in some metrics, and 8x better in raw compute. 9x more memory. “I am ...
  59. [59]
  60. [60]
  61. [61]
    Tesla HW5 Reveals Significant Performance Upgrade For New ...
    Aug 14, 2025 · Tesla has confirmed AI5 (hardware 5) FSD computer will be magnitude times more powerful than AI4 with improved camera sensors, ...
  62. [62]
  63. [63]
    Tesla Q&A Recap: Updates on HW3 Upgrade, FSD Unsupervised ...
    Jul 23, 2025 · AI5 / HW5 · Tesla wants to get FSD Unsupervised finished on AI4 before figuring out what to do with HW3 vehicles. · AI5 is still planned for next ...
  64. [64]
    Tesla AI5 & AI6 Chips "Compressing Reality"?! What Did Elon See?!
    Sep 14, 2025 · Tesla's upcoming AI5 chip will offer a 40x improvement over AI4, enabling significant advancements in Full Self-Driving (FSD) quality, safety, ...
  65. [65]
  66. [66]
    Owners Fully Engaged with Tesla's Traffic-Aware Cruise Control
    Apr 15, 2015 · Back in January 2015, Tesla Motors released its traffic-aware cruise control (TACC) feature for cars with Autopilot hardware.
  67. [67]
    Differences Between Tesla Basic Autopilot, Enhanced Autopilot ...
    Oct 26, 2024 · Autopilot consists of two key features, “Traffic-Aware Cruise Control” and “Autosteer”. Alongside these features, Tesla offers a whole host of ...
  68. [68]
    No auto steer on Tesla with 4/19 build date
    Sep 8, 2020 · All new Tesla cars come standard with Autopilot as of April 2019, which includes Traffic-Aware Cruise Control and Autosteer.
  69. [69]
    Does Tesla ever update AutoPilot software? - Quora
    Jun 28, 2023 · Yes. It has improved significantly since I got my Model 3 SR+ in July 2019 without FSD. Less phantom braking, smoother distance keeping.
  70. [70]
    (Mis-)use of standard Autopilot and Full Self-Driving (FSD) Beta - NIH
    Tesla's Full Self-Driving Beta (FSD) program introduces technology that extends the operational design domain of standard Autopilot from highways to urban roads ...Missing: rebranded | Show results with:rebranded
  71. [71]
    About Autopilot - Tesla
    Autopilot refers to a suite of advanced driver assistance features that are intended to make driving more convenient and less stressful.
  72. [72]
    About Autopilot - Tesla
    The features included with Full Self-Driving Capability are hands-on features. Keep your hands on the steering wheel at all times and be mindful of road ...<|separator|>
  73. [73]
    Navigate on Autopilot - Tesla
    Learn how to enable Navigate on Autopilot with either the Enhanced Autopilot or Full Self-Driving Capability package.
  74. [74]
    Tesla starts releasing Navigate on Autopilot feature with ... - Electrek
    Oct 24, 2018 · According to sources familiar with Tesla's software, the automaker slowly started releasing a new v9 update (2018.42) last night to the first ...<|separator|>
  75. [75]
    Introducing a More Seamless Navigate on Autopilot - Tesla
    Apr 3, 2019 · Introducing a More Seamless Navigate on Autopilot. The Tesla Team, April 3, 2019. Since we first introduced Navigate on Autopilot last year ...
  76. [76]
    Autopark - Tesla
    Autopark uses data to simplify parking on public roads by maneuvering Model 3 into parallel and perpendicular parking spaces.<|separator|>
  77. [77]
    Smart Summon - Tesla
    Smart Summon maneuvers Model Y out of parking spaces and around corners. This is useful for moving Model Y out of a tight parking spot, through puddles, or ...
  78. [78]
    Tesla reintroduces Enhanced Autopilot in the US for $6k
    Enhanced Autopilot is priced at $6,000, exactly half of Tesla's FSD package. Longtime Tesla owners are no stranger to EAP and its functionality. Enhanced ...
  79. [79]
    Full Self-Driving (Supervised) Subscriptions | Tesla Support
    If your vehicle has Full Self-Driving computer 3.0 or above, plus Basic Autopilot or Enhanced Autopilot, you are eligible to subscribe to FSD (Supervised).
  80. [80]
    Tesla reduces the price of FSD add-on, gets rid of Enhanced Autopilot
    Apr 21, 2024 · Now, Tesla has also dropped the price of the one-time add-on FSD option to $8,000 from $12,000 in the U.S., as can be seen on the company's ...
  81. [81]
    Elon Musk Says Every New Tesla Can Drive Itself - WIRED
    Oct 19, 2016 · Tonight, at a press conference, he announced that every new Tesla will be fully capable of driving itself. After being upgraded with a suite of ...
  82. [82]
    Full Self-Driving (Supervised) | Tesla Support
    In Q2 2025, we recorded one crash for every 6.69 million miles driven in which drivers were using Autopilot technology.
  83. [83]
    Full Self-Driving (Supervised) - Tesla
    The Full Self-Driving computer installed in Model Y is designed to use this input, rapidly process neural networks, and make decisions to safely guide you to ...
  84. [84]
  85. [85]
  86. [86]
    Tesla FSD Cost and Price Increase History
    Sep 5, 2023 · On the first of September, Tesla lowered the price of FSD by 20% in the U.S., down to $12,000. This brought FSD back to the price it was back at ...
  87. [87]
    Tesla cuts price of Full Self-Driving software by a third to ... - Reuters
    Apr 21, 2024 · Tesla on Saturday slashed the price of its Full Self-Driving (FSD) driver assistant software to $8000 from $12000 in the United States, ...
  88. [88]
    Tesla cuts price FSD premium driver assistance option by half in U.S.
    Apr 12, 2024 · Marketed as its Full Self-Driving, or FSD, package, Tesla customers will now pay $99 per month, down from $199 previously. The price cut is ...Missing: history | Show results with:history
  89. [89]
    Tesla is overhauling its Full Self-Driving subscription for easier access
    Aug 30, 2025 · Owners can purchase the suite outright for $8,000, or they can subscribe to the program for $99 per month, an option it enabled a few years ago.
  90. [90]
    Does Full Self-Driving (FSD) Transfer When You Sell a Tesla?
    Sep 26, 2025 · Update 4/25/2025:Again, for new vehicles delivered on/after April 24, 2025, Tesla is allowing transfer of FSD if it's paid in full and the car ...
  91. [91]
    Tesla (TSLA) brings back free FSD transfers, here's how to get it, $2 ...
    Apr 28, 2025 · Tesla Model 3s ordered on or after 24th April 2025 are eligible for the 0% APR offer. Tesla hasn't mentioned an expiry date for the new 0% APR ...
  92. [92]
    Transferring full self driving to new model x? - Facebook
    Oct 2, 2025 · As of October 2, 2025, the most recent program (which started April 24, 2025) ended on September 30, 2025, so transfers are not currently ...
  93. [93]
  94. [94]
  95. [95]
  96. [96]
    Why LiDAR is Doomed - Volt Equity
    Jan 25, 2021 · In typical Musk fashion, Tesla is a bit of an outlier for developing autonomous cars that will not use LiDAR (Light Detection and Ranging) as ...Basic Arguments Against... · Reframing The Debate From... · Each Sensor Has Its Own...
  97. [97]
    Tesla announces transition to 'Tesla Vision' without radar, warns of ...
    May 25, 2021 · Beginning with deliveries in May 2021, Model 3 and Model Y vehicles built for the North American market will no longer be equipped with radar.
  98. [98]
    AI & Robotics | Tesla
    FSD Chip Hardware. Neural Networks. Apply cutting-edge research to train deep neural networks on problems ranging from perception to control.Missing: v11 | Show results with:v11
  99. [99]
    Replacing Ultrasonic Sensors with Tesla Vision
    Sep 17, 2025 · In 2022 we took the next step in Tesla Vision by removing ultrasonic sensors (USS) from Model 3 and Model Y for most global markets, followed by all Model S ...
  100. [100]
    Tesla's radar-less Autopilot performance improves in heavy rain
    Aug 19, 2021 · Autopilot is reacting much more confidently in even heavy rainfall, drastically improving from the heavy speed reductions that were encountered in previous ...
  101. [101]
    Empirical Analysis of Autonomous Vehicle's LiDAR Detection ... - NIH
    Mar 9, 2023 · Under bad weather conditions, such as rain, snow, and fog, LiDAR-detection performance is reduced. This effect has hardly been verified in actual road ...
  102. [102]
    How Tesla will transition from Modular to End-To-End Deep Learning
    Sep 15, 2023 · A Deep Neural Network for every block; An End to End Model, assembling these neural networks together. Currently, for Tesla: Perception uses ...
  103. [103]
    Tesla FSD v12 shifts away from 'rules-based' approach - Teslarati
    Sep 10, 2023 · Isaacson talks about Tesla's recent development of the upcoming FSD v12, which he and Tesla demonstrate has moved away from a “rules-based” approach.<|separator|>
  104. [104]
    Tesla's Neural Network Revolution: How Full Self-Driving Replaced ...
    Jun 24, 2025 · The training requires 70,000 GPU hours per complete cycle, processing over 1.5 petabytes of driving data collected from Tesla's global fleet of ...
  105. [105]
    The Great Divergence: How Tesla's FSD v12 is Reshaping the ...
    Aug 4, 2025 · The FSD v12 update marked the transition to a single, end-to-end neural network that directly translates raw camera inputs ("photons";) into ...
  106. [106]
  107. [107]
    Autonomous driving: Modular pipeline Vs. End-to-end and LLMs
    Mar 15, 2025 · Modular pipeline: This is the older and more traditional approach for creating autonomous driving systems, and it is not preferred today, mainly ...Missing: rule- | Show results with:rule-
  108. [108]
    Tesla's Deep Learning at Scale: Using Billions of Miles to Train ...
    May 6, 2019 · Tesla can train its neural networks on as much useful data as it can collect. This means the size of its training dataset will correlate with its overall ...<|separator|>
  109. [109]
    Tesla's near-infinite training data isn't a silver bullet - Understanding AI
    Sep 12, 2024 · If Tesla's software makes a mistake every mile, then 90 percent of disengagements will be safety-related and Tesla can treat a disengagement as ...<|separator|>
  110. [110]
    Tesla to roll out FSD v14 in ~6 weeks with 10X parameters and ...
    Aug 11, 2025 · Tesla to roll out FSD v14 in ~6 weeks with 10X parameters and exponentially better safety, says Musk - Tesla Oracle.
  111. [111]
    Tesla FSD v14: 10x performance boost, new features & imminent ...
    Aug 12, 2025 · In addition to a massive 10x parameter upgrade, the update brings numerous new features such as pothole detection, significantly better parking ...
  112. [112]
    Navigating Autonomous Driving: Comparing Modular, End-to-End ...
    Nov 3, 2024 · The Modular Approach breaks down the driving task into well-defined stages, providing transparency and flexibility. The End-to-End Approach ...
  113. [113]
  114. [114]
    Review Performance, efficiency, and cost analysis of wafer-scale AI ...
    Tesla claims that Dojo can achieve 1.3 ExaFLOPS of compute power in a single cabinet with its 1.1 TB/s of inter-tile bandwidth and 10 TB/s on-tile bi-sectional ...
  115. [115]
    Tesla's fleet has accumulated over 1.2 billion miles on Autopilot and ...
    Jul 17, 2018 · That data is often referred to as “shadow mode” mileage because the autonomous driving system runs in the background of the vehicle without ...
  116. [116]
    How Tesla's Dojo Supercomputer is Revolutionizing AI Training for ...
    Jun 18, 2025 · Dojo's custom hardware achieves roughly 1.5 petaflops per kilowatt (FP16). By comparison, state-of-the-art GPU clusters deliver about 0.8 ...
  117. [117]
    Tesla shuts down Dojo, the AI training supercomputer that Musk said ...
    Aug 7, 2025 · Dojo's lead, Peter Bannon, is leaving the company, and the remaining team members will be reassigned to other data center and compute projects ...
  118. [118]
    Tesla disbands ambitious Dojo supercomputer team, shifts compute ...
    Aug 7, 2025 · Tesla is disbanding its Dojo supercomputer team, leader is leaving the company. Tesla will increase reliance on external tech partners like AMD ...
  119. [119]
    FSD 12.5.6 (2024.32.30) brings End-to-End on Highway for all Tesla ...
    Oct 9, 2024 · The newest installment of the Full Self-Driving software update v12.5.6 (2024.32.30) brings “End-to-End on Highway” to the entire vehicle lineup.Missing: shift | Show results with:shift
  120. [120]
  121. [121]
    Tesla Lowers Autopilot and FSD Strike Forgiveness Window to 3.5 ...
    Sep 17, 2025 · According to Tesla's owner's manual for software version 2025.32, the forgiveness period for each Autopilot or FSD strike has been cut in half, ...
  122. [122]
    Tesla Reduces FSD Strike Forgiveness to 3.5 Days
    Sep 17, 2025 · According to Tesla's owner's manual for update 2025.32, the forgiveness period for each Autopilot or FSD strike has now been cut in half, ...Missing: tweaks | Show results with:tweaks<|separator|>
  123. [123]
    Tesla rolls out FSD (Supervised) v12.6 to HW3 vehicles on 2024's ...
    Tesla rolls out FSD (Supervised) v12.6 to HW3 vehicles on 2024's 11th hour ... It appears that Tesla's AI team burned the midnight oil in the final days of 2024, ...
  124. [124]
    Tesla FSD V13 Review: New Features, Safety & V12 Comparison
    Apr 8, 2025 · FSD V13 has a stronger neural network, FSD activation from parked position, auto-parking, and improved speed profiles for smoother driving.
  125. [125]
    Musk Shares Tesla FSD Roadmap: What's in the Next FSD Update
    Other improvements listed in the FSD v13 changelog were a 3x context length improvement, audio inputs, improvements to reward predictions and false braking, and ...
  126. [126]
    FSD v13.2.* (HW4, Nov 2024 thru Sept 2025) - Tesla Owners Online
    Dec 3, 2024 · Integrated unpark, reverse, and park capabilities; Improved reward predictions for collision avoidance; Improved camera cleaning; Redesigned ...<|separator|>
  127. [127]
    Update 2025.32.8.16 (FSD 14.1.4) - Release Notes - Not a Tesla App
    Tesla software update 2025.32.8.16 includes FSD (Supervised) v14.1.4, Arrival Options, Speed Profiles, UI Improvements, Brake Confirm, Full Self-Driving ...
  128. [128]
    Tesla Reports 362-Mile FSD Drive With Zero Intervention - EV
    Aug 13, 2025 · Tesla announced on Tuesday that its Full Self-Driving (FSD) system was able to drive 362 miles (582 km) without requiring any intervention.
  129. [129]
  130. [130]
    Tesla's 2025 FSD Roadmap: Next FSD Update ... - Not a Tesla App
    Jul 28, 2025 · Tesla stated that it expects to launch FSD Unsupervised to certain US cities by the end of 2025. The key points here are select US cities, ...
  131. [131]
    Tesla's Full Self-Driving software under investigation for traffic safety ...
    Oct 9, 2025 · The federal safety agency in April 2024 closed an investigation into Tesla's less-capable Autopilot system after identifying 13 fatal crashes ...Missing: controversies | Show results with:controversies
  132. [132]
    Tesla's self-driving tech keeps being investigated for safety ... - CNN
    Oct 13, 2025 · Federal safety regulators are once again looking into Tesla's self-driving mode, the latest in a seemingly endless stream of investigations ...Missing: 2019 | Show results with:2019
  133. [133]
  134. [134]
    Tesla's Cybercab robotaxi is finally here with a $30K price tag
    Oct 11, 2024 · Musk said he expects the small EV to cost less than $30,000 when it arrives sometime before 2027. The Cybercab will charge via wireless ...
  135. [135]
    Robotaxi - Tesla
    Learn more about how Tesla is focused on creating a sustainable future by producing a fleet of autonomous vehicles for the Robotaxi network.
  136. [136]
    The Economic and Technological Potential of Tesla's Robo Taxi ...
    Mar 12, 2025 · This research paper examines the potential of Tesla's robo taxi business as a transformative force in the global transportation sector.
  137. [137]
  138. [138]
    How to Use Robotaxi | Tesla Support
    To request a ride with Robotaxi, follow these steps: Open the Robotaxi app. Enter and confirm your destination. Once confirmed, you will receive a notification ...
  139. [139]
    Get Started With Robotaxi | Tesla Support
    Learn more about how Robotaxi works, including scheduling a ride, how pricing works and where we currently provide service.
  140. [140]
    What We Know About Tesla's Robotaxi Launch - Business Insider
    Jun 10, 2025 · Musk has said he plans to expand the robotaxis to other cities, like San Francisco, following the Austin rollout, although he hasn't given an ...
  141. [141]
  142. [142]
  143. [143]
    Tesla's robotaxi launch could drive industry disruption and market ...
    Jun 25, 2025 · Tesla's (TSLA) robotaxi could reshape the auto market, Piper Sandler said. In a new note to clients, the firm reiterated its bullish stance.
  144. [144]
    Tesla Has Launched Its Robotaxi…Now What? - Ark Invest
    Aug 19, 2025 · According to our research, Tesla's robotaxi business could represent ~90% of its enterprise value by 2029, capturing a significant share of ...
  145. [145]
  146. [146]
    Tesla Q2 2025 vehicle safety report proves FSD makes ... - Teslarati
    Jul 23, 2025 · In Q2 2025, we recorded one crash for every 6.69 million miles driven in which drivers were using Autopilot technology. By comparison, the most ...<|separator|>
  147. [147]
  148. [148]
    Tesla Autopilot, FSD and Robotaxi Safety Versus Human
    Jul 8, 2024 · Tesla operation on highways with Autopilot are already eight times safer than human. Tesla recorded one crash for every 7.63 million miles driven with ...
  149. [149]
    A matched case-control analysis of autonomous vs human-driven ...
    Jun 18, 2024 · For example, Tesla Autopilot consistently maintains a closer distance to the lane center than human drivers. Additionally, Level 3 or ...
  150. [150]
    [PDF] Early Estimate of Motor Vehicle Traffic Fatalities for the First Quarter ...
    The fatality rate for the first quarter of 2025 decreased to 1.05 fatalities per 100 million VMT, down from the projected rate of 1.13 fatalities per 100 ...
  151. [151]
    Tesla Autopilot Was Uniquely Risky—and May Still Be | WIRED
    Apr 26, 2024 · A federal report published today found that Tesla's Autopilot system was involved in at least 13 fatal crashes in which drivers misused the system.
  152. [152]
    US probes Tesla's Full Self-Driving software in 2.4 mln cars after ...
    Oct 18, 2024 · NHTSA said it was opening the inquiry after four reports of crashes where FSD was engaged during reduced roadway visibility like sun glare, fog, ...
  153. [153]
    Tesla Fixes Another Recall With a Software Update - Not a Tesla App
    Jul 31, 2024 · This recall covers software failing to detect an unlatched hood before driving. Vehicles began receiving the software update in June to correct the issue.
  154. [154]
    Tesla FSD v14 Out - First Impressions - AutoPilot Review
    October, 2023 FSD Beta V11.4.7.3 (2023.27.7) The Tesla FSD Beta v11.4.7.3 update primarily focuses on bug fixes. The release notes for this version ...
  155. [155]
  156. [156]
    Tesla Deaths: Every Tesla Accident Resulting in Death
    We provide an updated record of Tesla fatalities and Tesla accident deaths that have been reported and as much related crash data as possible (e.g. location of ...
  157. [157]
    Do Teslas Struggle When the Sun Is Low? - InsideEVs
    Aug 8, 2025 · Beyond this TikTok clip, Tesla owners have reported FSD failures attributed to glare and sunlight. On Reddit, one owner described FSD ...
  158. [158]
    Musk Updates Tesla's FSD V14 Roadmap, Sets Timelines for V14.2 ...
    With FSD V14.1 rolling out, Elon Musk has revealed a multi-stage roadmap for V14.2 and a sentient V14.3, both expected in the coming weeks pending safety ...
  159. [159]
  160. [160]
    AMCI Testing releases its 2nd set of Tesla “Full Self Driving ...
    Oct 2, 2024 · As stated in our previous release, FSD's behavior required 75 interventions in 1000 miles ... rate AMCI testing experienced in only 1000 miles ...
  161. [161]
    Update Vehicle Firmware to Prevent Driver Misuse of Autosteer - Tesla
    Tesla has voluntarily issued a recall on certain model year 2012-2023 Model S that are equipped with Autosteer and were produced between October 5, 2012, and ...Missing: AP1 | Show results with:AP1
  162. [162]
    Tesla's software fixes, the NHTSA's status quo, and an impending ...
    Feb 28, 2022 · Software-based over-the-air fixes be dubbed and classified with the same terminologies as physical recalls, which typically involve the replacement of vehicle ...
  163. [163]
    Tesla starts using cabin cameras for driver monitoring - CNBC
    May 28, 2021 · "The cabin camera above your rearview mirror can now detect and alert driver inattentiveness while Autopilot is engaged. Camera data does not ...Missing: interior rollout date
  164. [164]
    Tesla's camera-based driver monitoring system rolls out to radar ...
    Sep 1, 2021 · Tesla has started rolling out its camera-based driver monitoring system to its radar-equipped vehicles as part of software update 2021.32.5.
  165. [165]
    [PDF] EFFECTIVENESS OF DRIVER MONITORING SYSTEMS
    In contrast, the driver monitoring system of the Tesla Model 3 provided consistent alert timings of approximately 37 to 39 seconds for both lighting conditions ...
  166. [166]
    Tesla Just Changed How It Punishes Inattentive Drivers - Autoblog
    Sep 26, 2025 · Software update 2025.32.100 comes with several revisions, most notably how it handles “strike forgiveness.” Tesla drivers accumulate “strikes” ...
  167. [167]
    Tesla Autopilot linked to hundreds of collisions, has 'critical safety ...
    Apr 26, 2024 · Tesla's Autopilot design has “led to foreseeable misuse and avoidable crashes,” the NHTSA report said. The system did not “sufficiently ensure ...
  168. [168]
    The Bigger Picture on Autopilot Safety - Tesla
    Jan 9, 2024 · Autopilot is ~10X safer than US average and ~5X safer than a Tesla with no AP tech enabled. More detailed information will be publicly available in the near ...
  169. [169]
  170. [170]
    Why You Should Listen To But Not Believe Tesla Naysayers
    Oct 23, 2024 · ... Peter Rawlinson's prediction that Tesla will not solve all of its FSD problem for at least another decade. This is in contrast to Tesla ...<|separator|>
  171. [171]
    Deep Dive: Tesla, Waymo, and the Great Sensor Debate
    Jul 8, 2025 · Tesla Autopilot data is also primarily from highway driving, where ... simulated miles on top of its millions of real-world driven miles.
  172. [172]
    Waymo vs Tesla: Who is closer to Level 5 Autonomous Driving?
    Sep 10, 2025 · Both seem to chase the same prize of "Level 5" autonomy, but when you look closer, their paths are so distinct they're barely comparable.Missing: criticism | Show results with:criticism
  173. [173]
    Skeptics of FSD are the only ones who haven't tried it, Elon Musk on ...
    Jan 31, 2025 · Elon Musk focused on the automaker's Autopilot Full Self-Driving (FSD) technology for the most part of the opening remarks of the Q4 2024 Tesla earnings call.
  174. [174]
    Tesla FSD turns off more U.S. consumers than attracts, survey finds
    Aug 28, 2025 · Only 14% of those surveyed said FSD would make them more likely to buy a Tesla, while 35% said the technology would make them less likely to ...
  175. [175]
    [PDF] INOA-RQ24009-12046.pdf - ODI RESUME
    Apr 25, 2024 · On December 12, 2023, Tesla filed a Defect Information Report (Recall 23V838) applicable to all. Tesla models produced and equipped with any ...Missing: details | Show results with:details
  176. [176]
    U.S. opens probe into 2.6 million Tesla vehicles over remote driving ...
    Jan 7, 2025 · The NHTSA probe covers 2016-2025 Model S and X vehicles, 2017-2025 Model 3 and 2020-2025 Model Y with the optional Full Self-Driving driver ...
  177. [177]
    Tesla drivers can pursue class action over self-driving claims, judge ...
    Aug 19, 2025 · Tesla must face a certified class action by California drivers who said Elon Musk misled them for eight years about the self-driving ...
  178. [178]
    Has Musk lied about self-driving Teslas? California says so
    Jul 24, 2025 · California authorities allege Tesla engaged in deceptive marketing by branding its technology as “Autopilot” and “Full Self-Driving Capability,” ...
  179. [179]
    Tesla Full Self-Driving Comes Out Of Beta, But Must Be Supervised
    Apr 1, 2024 · It appears that Tesla is officially taking "FSD Beta" out of beta, but only if you pay attention to the road.
  180. [180]
    A jury orders Tesla to pay more than $240 million in Autopilot crash
    Aug 2, 2025 · A Miami jury decided Tesla was partly responsible for a deadly 2019 crash in Florida involving its Autopilot driver assist technology.
  181. [181]
  182. [182]
    Top Fatal Autopilot Crash Lawsuit Settlements in California
    Settlement Details: A Miami jury awarded $329 million in total damages, including $200 million in punitive damages. Tesla was found 33% liable for the crash, ...Missing: outcomes | Show results with:outcomes
  183. [183]
    Digging Into Tesla's Liability in Crash Case: Where's the Data?
    Aug 6, 2025 · A Florida jury last week found that Tesla was 33 percent liable for an accident that killed a woman and injured her boyfriend in 2019. The two ...
  184. [184]
    NHTSA Adjusts Autonomous Vehicle Rules Ahead Of Tesla ...
    May 20, 2025 · On April 24, 2025, NHTSA issued new guidelines that grant the same exemption from FMVSS compliance to domestically manufactured autonomous ...
  185. [185]
  186. [186]
    Tesla FSD Europe: Hands-Free Highways LEGAL but STUCK?!
    Oct 2, 2025 · Tesla's full self-driving supervised allowing hands off but eyes on operation in europe and the uk which will make long drive safer.Missing: barriers | Show results with:barriers
  187. [187]
  188. [188]
    Tesla Eyes 2025 Supervised Full Self Driving Launch in China and ...
    Jul 28, 2025 · Tesla is preparing to expand its supervised driver assistance system in China and Europe next year, pending approval from local regulators.
  189. [189]
    Tesla halts driving-assistance software trial in China ... - Reuters
    Mar 24, 2025 · The company said last Monday that it would launch the free trial of its FSD service in China between March 17 and April 16. FSD is a suite of ...
  190. [190]
    Tesla working with Baidu to improve its smart driving system in ...
    Mar 13, 2025 · Baidu recently dispatched engineers to Tesla's Beijing office to work on better integrating its navigation map information with the FSD, ...
  191. [191]
    Tesla FSD improvements in China continues with Baidu's help
    Mar 13, 2025 · Tesla is working with Baidu to improve FSD in China, as strict data laws prevent Tesla from training its AI locally.
  192. [192]
    Tesla working with Baidu to improve assisted driving system in ...
    Mar 13, 2025 · Baidu engineers work with Tesla in Beijing to improve FSD navigation map - sources; Tesla delivered FSD V13 in China without sufficient local ...
  193. [193]
    Inside Tesla's Battle to Get Full Self-Driving Approved in Europe
    Sep 3, 2025 · The RDW documents cover a period from September 2024 to January 2025, and show that Tesla met weekly with regulators beginning in October.
  194. [194]
  195. [195]
  196. [196]
    Tesla Finally Reveals How Many Owners Are Buying FSD
    Sep 5, 2025 · For the first time, Tesla has revealed its FSD take rate, with a massive 50-60% of Model S/X buyers opting-in.
  197. [197]
    Tesla Says More Than 50% of Model S and X Drivers Pay for FSD
    Sep 3, 2025 · Tesla's VP of Vehicle Engineering says the FSD take rate is above 50 or 60% on Model S and X. Thanks in part to Jay Leno, we now have a better ...
  198. [198]
  199. [199]
  200. [200]
    Telsa Offers Insurance Discounts To Use Full Self-Driving
    Feb 4, 2025 · Use FSD 50% of the Time to Qualify​​ Tesla insurance customers can now qualify for a discount of up to 10% if they use the company's FSD system. ...Missing: Autopilot economic savings
  201. [201]
    Robotaxi Market Size And Share | Industry Report, 2030
    The global robotaxi market size was estimated at USD 1.95 billion in 2024 and is projected to reach USD 43.76 billion by 2030, growing at a CAGR of 73.5% ...
  202. [202]
    U.S. Robo Taxi Market Analysis Report 2025-2030, with Profiles of ...
    Aug 6, 2025 · The U.S. robo taxi market size was estimated at USD 0.45 billion in 2024 and is projected to grow at a CAGR of 74.6% from 2025 to 2030. The ...
  203. [203]
    Tesla Support - Full Self-Driving (Australia)
    Tesla's official support page for Full Self-Driving in Australia, detailing rollout and hardware requirements.