Tesla cites FSD safety stat around 1k

The fight here is about metrics — and about what kind of number should count as “proof” that Tesla is ready for wider robotaxi deployment. A bunch of recent posts pulled Tesla safety figures into the same argument and turned them into a simpler claim: about 1,000 miles on one metric, plus a four-second reaction window, therefore the system is already safer than what critics or regulators imply. But those numbers are coming from different buckets, and they do different jobs. (tesla.com) ### What number is Tesla actually publishing? Tesla’s current FSD safety page is a fleet-level collision page. It says vehicles using Full Self-Driving (Supervised) have fewer collisions than Teslas driven without it, and it frames the result as 7x fewer major collisions, 7x fewer minor collisions, and 5x fewer off-highway collisions. It also says the data covers more than 10 billion FSD miles, including about 3.8 billion city miles. That is a ver(tesla.com)fines them — not “miles between interventions” or “time-to-collision warnings.” (tesla.com) ### So where does the “around 1,000” idea come from? That number appears to come from a different genre of testing. In 2024, AMCI ran an independent 1,000-mile evaluation of Tesla FSD and said the system still made serious mistakes that would matter for robotaxi use. That was not Tesla’s official safety metric, and it was not a clean “1,000 miles between critical events” claim from Tesla. It was a third-party road test over roughly that distance. (tesla.com)they’re usually blending Tesla’s own safety framing with outside testing or crowdsourced intervention stats. (prnewswire.com) ### What about the four-second window? That seems to be another mix-up. Tesla’s Safety Score system uses forward-collision warnings and other driving-behavior factors to estimate future crash risk for insurance pricing and driver feedback. The material surfaced here do(prnewswire.com)k estimation — not a robotaxi-ready autonomy benchmark. (tesla.com) ### Why are people comparing Texas with California? Because Texas and California are operating under different political and regulatory vibes — and, in practice, different deployment stories. Austin has been the focal point for Tesla’s robotaxi push, while California has stricter supervised-testing structures and a more formal AV permitting culture. That makes Texas look friendlier, but it does not automatically mean Texas has (tesla.com)ving faster can be different. (robotaxi-safety-tracker.com) ### What are regulators actually worried about? The big federal question has been narrower than the online debate. NHTSA’s information request to Tesla focused on how a planned Austin robotaxi service would handle reduced-visibility conditions like rain, fog, dust, or glare, while an FSD defect probe was already open. In other words, the agency is not just asking whether aggregate fleet numbers look good. It wants to know how the system behaves in the edge cases that break autonomy systems. (static.nhtsa.gov) ### Why does this distinction matter so much? Because “safer on average” and “ready to run a driverless taxi fleet” are not the same claim. A supervised system can post good aggregate results while still having failure modes that are unacceptable without a fallback human. That’s basically the whole argument. Fleet collision rates help Tesla’s case. But intervention tests, edge-case failures, and regulator questions are aimed at a tougher standard. (tesla.com) ### Is the social-media version wrong, then? Not completely — but it is compressed past usefulness. Tesla really is publishing strong FSD safety claims. People really are using those claims to argue that Texas should be comfortable with bigger deployments. But the “around 1,000 miles” line and the “four-second window” line do not appear to be a single official Tesla safety benchmark. They are pieces from different measurement systems being snapped together into one talking point. (tesla.com) ### Bottom line? The live dispute is not whether Tesla has safety numbers. It does. The dispute is which numbers matter for robotaxis — broad collision averages, driver-warning proxies, or edge-case performance when nobody is sitting there to save the car. (tesla.com)