Fred Lambert. Deploy

Policy6 hours ago

Tesla FSD v14 is so good it’s making me dangerously complacent

Electrek's Fred Lambert, reporting from weeks of daily driving Tesla FSD v14, says the system now travels roughly 2,000 miles between critical interventions but argues Tesla's marketing (official X posts calling FSD a "personal chauffeur," promoting elderly and vision-impaired buyers, and Musk endorsing texting "depending on context") is actively encouraging driver complacency on what remains a Level 2 system. The piece cites a recent crash involving former Uber self-driving head Raffi Krikorian in a Tesla Model X on FSD, plus IIHS warnings, expanded NHTSA Level 2 crash reporting, and a Reuters report that seven of nine former Tesla data labelers wouldn't trust FSD to drive them.

Why it matters

The operator number to anchor on is 2,000 miles between critical interventions, which translates to 30 to 40 hours of hands-off-attention driving between failures. That is the worst possible regime for a Level 2 system: reliable enough to train the driver out of vigilance, unreliable enough that the handoff still arrives, and a documented 5 to 8 second reengagement lag when it does. The "ironies of automation" literature is forty years old; the new variable is that Tesla's official channels are now actively selling FSD to demographics (a 71-year-old with failing eyesight, a 93-year-old, a driver born without arms) who structurally cannot perform the supervisory role the EULA requires.

For anyone modeling Tesla's autonomy liability, those tweets are the exposure. They are durable, attributable to the corporate account, and aimed at exactly the buyer profile a plaintiff's attorney will reconstruct after a fatal handoff failure. NHTSA's expanded Level 2 crash reporting and the IIHS's standing recommendation to redesign driver monitoring give regulators the on-ramp. The Krikorian crash matters because it removes the "inattentive driver" defense from the population most equipped to supervise. The structural question for the category is whether Tesla's vision-only stack can close the gap to unsupervised L4 before the marketing-versus-classification mismatch produces a forcing event from NHTSA or a state AG.