TL;DR

Musk describes dozing off during a commute and arriving safely as the key test for Tesla autonomy. Austin robotaxis crash four times more frequently than human drivers, while 4 million HW3 vehicles lack the hardware for unsupervised operation.

Elon Musk has long described falling asleep in a Tesla and waking at the destination as the true measure of full autonomy. He outlined this idea in 2014 and revisited it during Tesla's Q1 2025 earnings discussion, expressing confidence in availability across many US cities by year end. That target was missed.

During the Q1 2026 earnings call in April, Musk revised expectations once more, stating that unsupervised Full Self-Driving for consumer cars would not appear before Q4 2026. He further noted that Hardware 3, found in about four million Teslas, lacks the power needed for unsupervised FSD.

Safety data and its limitations

Tesla's safety reports show one major incident every 5.3 million miles with FSD Supervised, versus one every 660,000 miles for average US drivers. The fleet reached 10 billion cumulative FSD miles in May 2026.

These figures cover FSD Supervised, a Level 2 feature requiring constant human attention. They cannot be directly compared to unsupervised systems. Tesla applies distinct crash reporting standards from NHTSA benchmarks, a difference safety experts have highlighted often.

Austin robotaxi performance details

The Austin driverless service serves as the main real-world example of unsupervised operation. Started in June 2025 with safety drivers, it moved to fully driverless trips in January 2026 and now spans 245 square miles in central Texas using around 20 vehicles.

Actual safety results differ from promotional claims. Tesla logged 14 crashes with NHTSA over about 800,000 Austin miles through February 2026, equaling one incident per 57,000 miles. This rate runs roughly four times higher than the human average referenced on Tesla's safety page.

Fleet expansion issues

As of late May 2026, Tesla had 42 vehicles cleared for driverless use in Texas, far below Waymo's 577. The active fleet has decreased, with expansion delayed until FSD v15 arrives, a step Musk ties to late 2026 or early 2027.

Statements about eight million Teslas running robotaxi-derived software hold only because FSD Supervised shares code with the robotaxi setup. These vehicles remain at Level 2 and need human oversight. Half, the four million with Hardware 3, cannot support unsupervised FSD even with updates.

Contract changes

From 2016 to early 2024, Tesla marketed FSD as Full Self-Driving Capability without the supervised label. The name shifted to Full Self-Driving (Supervised) in March 2024 with version 12.3.3. In June 2026, owners found Tesla had updated older purchase documents to add supervised wording, hiding the prior contract text.

Tesla faces a certified US class action over FSD advertising claims from October 2016 through August 2024. The European Transport Safety Council has recommended a cautious stance on the system due to risks of driver over-reliance.

Key concerns

The 5.3 million miles-per-crash number comes from Tesla's own reporting methods, which differ from NHTSA standards, and applies only to supervised use. Unsupervised data from Austin shows a crash rate about four times the human average.

Reports of a 52,000-mile FSD v14 trip without disengagements lack independent confirmation. Verified records list a 2,833-mile coast-to-coast journey as the longest confirmed stretch, with typical intervention-free distances near 25 miles.

Musk's sleeping-commute test has been promised since 2014. Current plans set unsupervised FSD for consumer vehicles no sooner than Q4 2026, limited to HW4 cars in approved zones. Four million HW3 owners who paid up to $15,000 for FSD will require upgrades Tesla has not yet scheduled or priced.