TSLAThursday, 3 Sep 2026

Tesla FSD: row over stale EU safety data, and the Cybercab launch event in Austin goes ahead

Today brought 2 catalysts, and both come with qualifications. (1) Independent media have picked apart the FSD safety dashboard that Tesla published on 1 Sep ahead of the EU vote on 6 Oct. All the data behind the 4.1x collision-probability advantage end in 2025 (the latest is 5 Dec), about 9 months behind the April to August 2026 period it claims to cover.

The European mileage totals also do not reconcile internally (755,922 against 793,000+). This walks back the positive framing of 2 Sep. It is a downgrade to the quality of delivery, not a negative big enough to move valuation. (2) Today's Cybercab launch event in Austin went ahead, giving the public its first rides in a Robotaxi with no steering wheel or pedals. But first-hand reporting from WIRED and The Verge confirms that a safety monitor still sits in the front passenger seat, so the threshold to unsupervised commercial operation has not been crossed. The event carries the 23 Aug date lock and the 45 registrations of 1 Sep through to the unveiling. It is an incremental catalyst, not a turning point in volume.

Key evidence

Stale FSD data: today Automotive World and TheStreet picked apart the FSD safety dashboard published on 1 Sep. It claims to cover April to August 2026, but the mileage breakdown is for the whole of 2025, the automatic emergency braking data run to 31 Oct 2025, the lane-change data stop in early Dec 2025, the latest observation is 5 Dec 2025 and there are no records at all for 2026.
Contradiction inside the FSD data: the European mileage table adds up to 755,922 miles, yet elsewhere in the same file Tesla cites a cumulative total of more than 793,000 miles across 8 countries, and it has not explained the difference in basis.
The EU vote: the EU-wide vote on 6 Oct still needs the backing of 15 of 27 member states. It has been pointed out that publishing a file whose latest data are 9 months old is not the transparency regulators want.
The Cybercab launch event in Austin took place today: Reuters, Teslarati, The Verge and WIRED described it as the first time the public can ride a Robotaxi on public roads in a vehicle with no manual controls at all.
WIRED stressed that in most Cybercab road tests so far, a safety monitor still sits in the front passenger seat, ready to take over at any time.
The Verge noted that for the millions of existing owners, FSD is still supervised L2, and that the Cybercab is the first to go unsupervised.

Show 3 more

Robotaxi in the San Francisco Bay Area is still not legally allowed to run without a driver.
17 Cybercabs were photographed in large-scale testing in Bellevue, Washington state (a new testing location).
This follows the lock-in of the 3 Sep date on 23 Aug and the 45 vehicles entering Texas commercial registration on 1 Sep. Today the event itself took place.

Catalysts to watch

EU-wide FSD vote on 6 Oct (it is unknown whether the stale-data row will affect the threshold of 15 of 27 votes)
Official figures after the Cybercab event to verify real operating scale and ride volumes
Removing the safety monitor and scaling up to paying public riders: the key threshold for genuinely unsupervised commercial operation

Risks

Stale data does not mean falsified data. The dashboard may simply be slow to update.
The 4.1x figure is Tesla's own and already needed a discount.
A launch event and invited test rides are not scaled-up service for paying members of the public, and they contain an element of staging.
No steering wheel is a hardware form. Having no one supervising is the business model. The same day's evidence shows the 2 are separate.

Sources 6

Automotive World (3 Sep 2026)
TheStreet (3 Sep 2026)
Reuters (3 Sep 2026)
Teslarati (3 Sep 2026)
The Verge (3 Sep 2026)
WIRED (3 Sep 2026)

A brief is a lead for our research, not a verified conclusion or advice.