Tesla has begun pushing another Full Self-Driving software patch to vehicles across North America, continuing the rapid update cycle behind one of the automotive industry’s most closely watched driver-assistance systems. Software version 2026.38.5.11 began appearing in fleet-tracking data on October 24, carrying Full Self-Driving (Supervised) v14.3.13 on newer AI4-equipped vehicles and a lighter v14.3 build on older HW3 cars.
The update is incremental rather than a completely new FSD generation, but its timing makes it significant. Tesla is expanding FSD adoption while federal regulators continue examining alleged traffic-law violations and performance in reduced-visibility conditions. At the same time, Tesla is publishing increasingly detailed safety statistics arguing that supervised FSD reduces collision risk. The result is a software rollout unfolding alongside a much larger debate over how quickly increasingly capable driving automation should advance—and how closely humans must continue watching it.
A Fresh Patch Is Expanding Across North America
Tesla software version 2026.38.5.11 began reaching vehicles on October 24, with additional installations appearing through October 25. Independent Tesla fleet trackers characterize the release as a North American rollout and show it reaching vehicles from several model lines. As with most Tesla over-the-air updates, distribution is staged rather than simultaneous. That means two otherwise similar Model Ys parked in the same city may remain on different builds for days while Tesla expands deployment and watches fleet performance. The company generally uses progressive software rollouts so updates can reach increasingly large groups without requiring dealership visits.
What makes this particular release easy to misunderstand is that 2026.38.5.11 is not introducing a newly numbered FSD version for AI4 vehicles. It continues to carry FSD (Supervised) v14.3.13, which was already present in the preceding 2026.38.5.10 build. Published release notes explicitly indicate that the v14.3.13 notes remain unchanged. In practical terms, that makes the latest software better described as a patch or maintenance release than a major autonomous-driving upgrade. Tesla does not publicly provide a complete change log for every underlying fix, so specific improvements unique to the “.11” patch should not be assumed unless the company documents them.
What FSD v14.3.13 Is Designed to Improve
Although the latest point release does not bring a fresh FSD feature list, the v14.3.13 software it carries contains a substantial collection of changes. Tesla’s release notes describe an upgraded reinforcement-learning training approach, improvements to its vision encoder for rare and reduced-visibility situations, stronger recognition of traffic signs and three-dimensional road geometry, and better handling of complicated traffic lights. Tesla also says the software improves responses involving emergency vehicles, school buses, right-of-way violations, unusual road objects and small animals. Another engineering change is claimed to reduce reaction time by roughly 20 percent through changes to the software compilation pipeline.
Several improvements concern situations that matter more than headline-grabbing demonstrations of a car navigating city streets. Tesla says v14.3.13 is better at detecting temporary system degradation and recovering when conditions improve. Driver-monitoring sensitivity has also been revised, including eye-gaze estimation, recognition through different eyewear and performance under changing cabin lighting. Other adjustments target lane-position bias, following distance, parking and stopping behavior around yellow lights. Those details matter because successful driver-assistance software is often judged less by spectacular maneuvers than by how predictably it handles thousands of ordinary edge cases. Still, these are Tesla’s stated improvements to v14.3.13—not evidence that 2026.38.5.11 itself has independently solved each of those problems.
HW3 and HW4 Cars Are Not Running the Same FSD Stack
One of the most consequential details in the rollout is the growing distinction between older Hardware 3 vehicles and newer cars equipped with Tesla’s AI4, commonly called HW4. AI4 vehicles receiving the update run FSD v14.3.13, while supported HW3 cars receive what Tesla calls FSD v14.3 Lite. That distinction reflects the different computing resources available in the two hardware generations and means an owner seeing “v14” on an HW3 vehicle should not assume the car is running exactly the same neural-network stack as a recently built Model Y with AI4.
Tesla says v14.3 Lite uses intelligence distilled from the newer HW4 version while being optimized for the earlier computer. Its documented capabilities include improved responsiveness and driving comfort, parking and unparking, reversing when necessary, arrival choices and adjustable speed profiles. The strategy allows Tesla to continue bringing parts of its newer FSD development to a large installed base without pretending that the hardware is identical. For owners, however, it creates a more fragmented software landscape: the vehicle model alone no longer tells the whole story. Two Model 3s built several years apart can display similar FSD branding while operating substantially different underlying software because of their onboard computers.
Driver Monitoring Is Becoming More Central to the Experience
As FSD becomes more capable, Tesla is simultaneously making the system more attentive to what the person behind the wheel is doing. The company’s current manuals describe a cabin camera that monitors attentiveness while FSD is operating. Drivers who repeatedly look away can receive escalating warnings, and Tesla can disable the feature for the remainder of a drive when those warnings are ignored. Under some circumstances the vehicle can request steering-wheel input as another confirmation of engagement. Tesla also warns against attempts to circumvent its monitoring systems and says a vehicle can activate warnings and bring itself to a stop when the driver fails to respond appropriately.
There is a human-factors reason for those safeguards. A 2025 Insurance Institute for Highway Safety study followed 14 people using Tesla Autopilot—not the current FSD v14.3.13 software—and found that drivers learned how the system’s attention reminders worked over time. Participants became faster at responding to warnings while also increasing some off-road glances and non-driving activities. The finding illustrates a difficult problem across partially automated vehicles: better automation can make supervision feel less demanding precisely when supervision remains essential. The National Transportation Safety Board has consequently pushed for stronger driver-monitoring standards capable of detecting accumulated distraction, phone use and other signs of disengagement. Tesla’s enhanced gaze monitoring therefore sits at the centre of the safety case, not on its margins.
Federal Investigations Keep the Safety Debate Open
Tesla’s latest rollout is occurring while the U.S. National Highway Traffic Safety Administration continues examining separate safety issues associated with FSD. One investigation, opened in October 2025, concerns alleged traffic-law violations while FSD was engaged. The agency’s opening document highlighted reports involving red-light violations, improper lane choices and vehicles allegedly entering opposing traffic. At that stage, regulators identified 18 complaints concerning red lights and six crash reports submitted through NHTSA’s Standing General Order process, four of which involved injuries. A subsequent information request in December said the agency had identified a larger body of potentially relevant material, including 62 complaints and 14 Standing General Order reports.
A second investigation was advanced to an engineering analysis in March 2026 and focuses on performance during reduced visibility or situations in which camera performance may be degraded. NHTSA said it was examining incidents involving conditions such as glare and airborne obscurants and whether the system appropriately detects degradation and warns the driver. Reuters reported that the analysis covers approximately 3.2 million vehicles and involves nine incidents under review, including one fatal crash and two injury crashes. An investigation is not a finding that Tesla has a safety defect, and it does not automatically result in a recall. It is nevertheless notable that v14.3.13 specifically claims improvements involving low visibility, traffic lights and degradation detection. Tesla has not publicly characterized 2026.38.5.11 as a regulatory remedy for either investigation.
Tesla’s Safety Numbers Are Strong on Paper—and Still Contested
Tesla argues that the enormous volume of real-world telemetry collected by its vehicles provides evidence that FSD (Supervised) improves safety. Its current Vehicle Safety Report says vehicles using FSD experience seven times fewer major and minor collisions and five times fewer off-highway collisions than the comparisons used by Tesla. The company defines collision events using measurable thresholds such as airbag or other non-reversible restraint deployment and changes in vehicle velocity. It also counts a crash as involving FSD if the system was active at any point during the five seconds before the collision, preventing a last-second disengagement from automatically removing the event from the FSD category.
The dispute concerns how those numbers should be interpreted. Tesla says comparisons between FSD-driven Teslas and manually driven Teslas are particularly valuable because the vehicles share a common fleet and telemetry pipeline. Its methodology also acknowledges differences between its data and national crash statistics. Reuters, however, reported criticism from outside researchers who argued that some broader Tesla-versus-U.S.-average comparisons can be distorted by different crash-reporting thresholds, fleet age and vehicle safety characteristics. In June, U.S. Senators Edward Markey and Richard Blumenthal asked NHTSA to review Tesla’s published FSD safety claims following that reporting. The debate does not make Tesla’s telemetry meaningless; it demonstrates why standardized definitions, independently reproducible analysis and equivalent comparison groups become increasingly important as safety statistics influence public acceptance and regulation.
Canadian Drivers Face the Same Supervision Rules
For Canadian Tesla owners, the “Supervised” part of the FSD name remains just as important as the software version. Tesla’s Canadian support materials say FSD can navigate roads, make turns, change lanes and negotiate intersections but does not make the vehicle autonomous. The driver must remain attentive and responsible for the vehicle. Tesla Canada currently lists FSD as a $99-per-month subscription and says the system can be engaged when the vehicle is travelling below 150 km/h, subject to vehicle eligibility, software and other conditions. Those rules are unaffected simply because a newer patch has arrived over the air.
Transport Canada’s description of vehicle automation reinforces the same distinction. Under the widely used SAE framework, Level 2 partial automation can control steering as well as acceleration and braking, but the human driver must continue monitoring the road and responding to the environment. The federal department also warns that driver-assistance technologies may work poorly—or not at all—in conditions including rain, snow and some infrastructure such as tunnels. That qualification has particular relevance in Canada, where winter weather can obscure lane markings, coat cameras and dramatically change road surfaces within a single trip. A smoother FSD update can reduce workload, but it does not transfer legal or practical responsibility away from the person in the driver’s seat.
Why a Small Patch Matters to Tesla’s Bigger Autonomy Strategy
The commercial stakes behind seemingly routine FSD patches have become much larger. Tesla reported 1.74 million “active FSD subscriptions” at the end of the third quarter of 2026, up from 1.04 million a year earlier. Importantly, Tesla’s corporate definition includes both customers who previously paid upfront and those using monthly subscriptions, while excluding free trials. The company also said more than 75 percent of new vehicle deliveries in North America during the quarter included a sign-up for a paid FSD subscription at delivery, nearly 20 percentage points higher sequentially. Tesla’s finance chief clarified that customers were making those sign-ups even though newly delivered cars also receive a free trial.
Tesla is already discussing the next stage. During its October earnings call, AI chief Ashok Elluswamy said software related to the company’s Cybercab and Model Y programs could reach customers as part of v15 “pretty soon,” and suggested some customer vehicles could potentially join Tesla’s Robotaxi fleet around the second quarter of 2027 if requirements such as geofencing and fleet management are satisfied. That remains a company forecast rather than a guaranteed deployment schedule. For now, 2026.38.5.11 represents the less dramatic reality of autonomy development: another staged patch, another round of fleet data and another test of whether improved capability can arrive without weakening the expectation that a human remains ready to intervene.
