Elon Musk is once again promising that the next version of Tesla’s “Full Self-Driving” software will “far exceed human levels of safety.” This time it’s V15. The problem is that he made virtually identical claims about Version 12 in 2023 and V14 in 2025 — and neither came close to delivering.
Musk posted the claim today in response to a user’s detailed review of FSD v14.3 after 600+ miles of testing. He acknowledged that “point releases will bring polish” to the current version before projecting that V15, the next major release, will achieve what every previous version was supposed to achieve.
The pattern: every version is ‘the one’
Here’s the timeline of Musk tying specific FSD version numbers to claims of superhuman safety, taken directly from his posts on X:
November 2023 — Version 12: Musk wrote that “Supervised FSD is vastly safer than human driving” and that “Version 12 (end-to-end neural nets) will far exceed human safety even when unsupervised.” Version 12 shipped. It did not far exceed human safety. It remained a supervised Level 2 system.
August 2025 — V14: Musk stated that “V14 will be better than human for sure, but I don’t know if it will be 10X. Maybe 2X to 3X. V15 has a shot at 10X.” Two days later, he added that V14, “while capable of being much safer than the average human driver, will take a few months to debug post release.” V14 shipped. It was an incremental improvement — still requiring constant supervision, still experiencing hallucinations and erratic behavior.
April 9, 2026 — V15: Musk posted today that “V15 will far exceed human levels of safety, even in completely unsupervised and complex situations.” The goalposts move to the next version number once more.
The language barely changes. “Far exceed human safety” in November 2023 becomes “better than human for sure” in August 2025, which becomes “far exceed human levels of safety” again today. Only the version number increments.
The data tells a different story
The gap between Musk’s rhetoric and Tesla’s actual safety performance remains enormous. As we reported yesterday, Musk’s claim that FSD is “10X safer” than humans doesn’t hold up under scrutiny. Tesla’s own safety comparisons suffer from fundamental methodological problems: road-type mismatch (FSD operates mostly on highways, the safest roads), vehicle-age mismatch (Teslas are among the newest cars on the road), driver demographic differences, and different crash definitions than NHTSA uses.
Tesla does not publish disengagement data, crash-severity breakdowns, or the methodology behind its safety claims. It has never released peer-reviewed safety research with matched human-driver baselines.
Waymo, by contrast, has published peer-reviewed data across 56.7 million rider-only miles showing an 85% reduction in injury crashes and 90% fewer serious-injury-or-worse crashes compared to human drivers on the same roads. That’s actual data — not a promise about the next software update.
Meanwhile, NHTSA has upgraded its investigation into FSD’s inability to handle reduced visibility conditions, expanding the probe to cover 3.2 million vehicles. The agency is also running a separate investigation into FSD traffic violations including running red lights and crossing into oncoming lanes. That’s three concurrent federal investigations into a system Musk insists is safer than you.
A decade of the same promise
This isn’t just a pattern across FSD version numbers. Musk has been moving the goalpost on Tesla’s self-driving timeline for over a decade. Full autonomy was promised by 2018. A coast-to-coast autonomous drive was coming in 2017. One million robotaxis were supposed to be on the road by 2020. Unsupervised FSD was definitely happening in June 2025.
None of it happened. FSD remains a Level 2 driver-assistance system that requires constant supervision. Tesla just launched FSD v14.3 this week with incremental improvements — a rewritten AI compiler and 20% faster reaction times — but it’s still supervised, still imperfect, and still far from the unsupervised autonomy Musk keeps promising is one version away.
As for V14 specifically — the version Musk said would be “better than human for sure” — FSD v14.3 just launched this week and Musk himself described it as needing “polish” via point releases. That doesn’t sound like a system that exceeded human safety.
Electrek’s Take
We’ve been covering Musk’s self-driving promises for years now, and the pattern is unmistakable: every major FSD version is “the one” that will finally beat human drivers — right up until it ships and doesn’t. Then the claim quietly migrates to the next version number.
Tesla has never released any data indicating that FSD is safer than human drivers. The only thing it did was release heavily massaged data that points to human drivers driving more safely with FSD than the average US driver without.
The specific version-safety claims are worth documenting because they show just how precisely Musk recycles the same language. “Far exceed human safety” was the promise for Version 12 in 2023. It’s now the promise for V15 in 2026. Nothing about the underlying evidence has changed — Tesla still publishes no peer-reviewed safety data, still faces multiple NHTSA investigations, and still sells a supervised Level 2 system that requires you to keep your hands on the wheel.
Meanwhile, Waymo is actually operating unsupervised robotaxis in multiple cities with published, peer-reviewed safety data showing dramatic safety improvements over human drivers. That’s the standard for what “far exceeding human safety” actually looks like — and it involves transparent data, not social media posts promising the next update will be the breakthrough.
We’re not saying Tesla’s FSD will never get there. But after a decade of identical promises attached to incrementing version numbers, the appropriate response to “V15 will far exceed human safety” is: show us the data when it ships. The track record of these specific claims is 0 for 3.
FTC: We use income earning auto affiliate links. More.
electrek.co
#Musk #Tesla #FSD #v15 #exceed #human #safety #v12 #v14
