Tesla is preparing a major safety upgrade for Full Self-Driving v15 that promises faster hazard detection and improved collision avoidance, but the company's opacity around crash data continues to fuel transparency concerns. The next version of Tesla's driver-assist software will feature a much larger neural network and tighter integration between prediction and control systems, according to Tesla's AI leadership.

What Safety Improvements Is Tesla Planning for FSD v15?

Tesla AI lead Ashok Elluswamy outlined the coming improvements after a near-miss incident where a Tesla driver's car swerved to avoid another vehicle pulling out of a parking lot. Elluswamy stated that the company would deliver "even earlier prediction of hazards, even faster reaction time and overall significantly better safety and collision avoidance" with the v15 update. The company has already begun testing early v15 software in some robotaxi operations while rolling out intermediate safety features like Automatic Collision Evasion into current customer vehicles, allowing the driving stack to intervene even when drivers are manually controlling the car.

Elluswamy

The architectural redesign represents Tesla's continued strategy of treating software as the primary path to safety improvements. By expanding the neural network and creating tighter coupling between the system's prediction and control layers, Tesla aims to reduce the gap between what the software perceives and how quickly it can respond to hazards.

How Does Tesla's Safety Data Compare to Human Drivers?

Tesla publishes telemetry data to support its safety claims. According to the company's North American Vehicle Safety Report, vehicles with FSD (Supervised) engaged traveled between 5.1 million and 5.7 million miles between major collisions, defined as airbag-deployment events. Tesla estimates the U.S. average for comparable crashes is about 699,000 miles, which the company frames as roughly seven times fewer major collisions. However, independent researchers argue a more accurate comparison uses Tesla's own fleet baseline: cars driven manually with active safety features like automatic emergency braking recorded a major collision about every 2.1 million miles. Against that baseline, FSD's advantage shrinks to roughly 2.4 to 2.7 times fewer severe crashes.

European data released in 2026 pointed in a similar direction, with Tesla reporting FSD as 3.5 times safer than manual driving in the Netherlands and 4.1 times fewer collisions than manually driven Teslas with active safety across more than 100 million kilometers in five approved countries.

Why Are Transparency Concerns Overshadowing Safety Claims?

Despite these statistics, significant questions persist about how much the public actually knows about FSD's real-world performance. Under a 2021 National Highway Traffic Safety Administration (NHTSA) Standing General Order, automakers must report crashes where a Level 2 driver-assist system was engaged within 30 seconds of impact and someone was hurt, killed, or a vehicle was towed. Tesla accounts for roughly 85 percent of the entire industry's reports, with nearly 4,000 crashes documented, and the count climbs every year.

The critical problem lies in what Tesla does with those reports. On virtually all of them, the company redacts the crash narrative, the software version, and the field indicating whether the road was within the system's approved operating area, all as "confidential business information." Tesla has blacked out the narrative on 99.9 percent of its reports every single year since 2019. Almost no other automaker does this; General Motors, Ford, Honda, and Toyota redact essentially none of their narratives.

What Do Investigators Discover When They Match Redacted Reports to Public Records?

Researchers have begun matching Tesla's redacted NHTSA reports to specific fatal crashes covered in local news to understand what the company is hiding. By converting UTC timestamps to local time and cross-referencing city, vehicle type, and fatal outcomes, investigators have confirmed several cases where FSD or Autopilot was verified engaged but never mentioned in public accounts of the crashes.

One striking example involved a 23-year-old IT technician named Steven Alvarez in Clute, Texas, whose 2025 Model 3 crashed at 104 miles per hour on May 20, 2026. Local police reported their preliminary belief was that Alvarez had suffered a medical episode, and no outlet mentioned Tesla's driver-assist system. However, Tesla's filing to NHTSA marked the automation "Engagement Status" as "Verified Engaged," meaning Tesla confirmed through its own telematics that Autopilot or Full Self-Driving was active. The 104 miles per hour speed figure exists only in Tesla's telematics data; it appeared nowhere in the public record.

Additional confirmed cases include:

• Lake Mary, Florida: A 43-year-old man died when his Model 3 sat stopped in a live Interstate 4 lane and a semi rear-ended it; Tesla's data shows driver-assist verified engaged at 0 miles per hour.
• Mesa, Arizona: A near-identical case with another 2020 Model 3 stopped at 0 miles per hour on Loop 202 at 3 a.m., rear-ended, with driver-assist verified engaged.
• Batavia, Illinois: Maggie Espinosa, 37, a teacher and mother of five, was killed as a passenger when a Model Y turning left at an intersection struck a pickup; Tesla's data shows driver-assist verified engaged while making the turn, a maneuver pointing to Full Self-Driving.
• Buena Vista Township, New Jersey: A Tesla Model 3 ran a stop sign and killed 82-year-old Stephen Field; Tesla's data shows driver-assist verified engaged at 4 miles per hour, a rolling-stop signature.

How to Understand the Gap Between Tesla's Marketing and Its Technology?

Several key distinctions help explain why transparency matters so much in this debate:

• Legal Classification: Full Self-Driving remains a Level 2 driver-assist system, meaning the human driver is legally responsible at all times, even though Tesla markets it as "Full Self-Driving" and Elon Musk repeatedly promises unsupervised autonomy is months away.
• Complacency Risk: The gap between marketing and technical reality breeds complacency; drivers who over-trust the system may stop paying attention, increasing crash risk when the system reaches its limits.
• Data Asymmetry: Tesla holds the telematics, video, software version, and driver-monitoring data that would explain what happened in fatal crashes, but the public is denied access to this information while being encouraged to trust the marketing.

The company justifies its redactions by claiming it would "suffer financial harm" if the information became public. Critics argue this reveals a troubling priority: Tesla is telling regulators that the commercial value of its secrecy outweighs the public's right to know how its cars behave in fatal crashes.

As Tesla prepares to roll out v15 with promises of better safety, the fundamental tension remains unresolved. The company is simultaneously pushing owners to believe the car can handle driving itself while hiding the records that would show how the system actually behaves when things go wrong.

"Tesla is blacking out the single most important fields in a federal safety database, what its car did, which software was running, and whether the car was even supposed to be operating that way, in reports about crashes that killed people," according to investigative reporting on the issue.

Electrek Investigation

Until that transparency gap closes, even significant technical improvements may struggle to rebuild public trust in the system.