Tesla FSD “acid test” vs Austin robotaxi crash reality



TL;DR

Musk calls sleeping through your commute the “acid test” for Tesla FSD. His Austin robotaxis crash four times more often than human drivers, and 4 million HW3 Teslas cannot achieve unsupervised driving at all.

Elon Musk has called the ability to fall asleep in a Tesla and wake up at your destination the “acid test” for true autonomy. He first described that vision in 2014 and repeated it on Tesla’s Q1 2025 earnings call, saying he was confident it would be available in many US cities by the end of that year. It was not.

On the Q1 2026 earnings call in April, Musk pushed the timeline again, saying unsupervised Full Self-Driving for consumer vehicles would not arrive until Q4 2026 at the earliest. He also admitted that Hardware 3, the computing platform inside roughly four million Teslas on the road, “simply does not have the capability to achieve unsupervised FSD.”

The safety statistics and what they leave out

Tesla’s safety page reports one major collision per 5.3 million miles under FSD Supervised, compared with one per 660,000 miles for the average US driver. The fleet crossed 10 billion cumulative FSD miles in May 2026.

Those numbers describe FSD Supervised, the Level 2 system where a human driver must remain attentive and ready to take over at all times. They are not comparable to unsupervised driving. Tesla uses different crash-counting methods than the NHTSA data it benchmarks against, a discrepancy that safety researchers have flagged repeatedly.

What the Austin robotaxi data actually shows

Tesla’s driverless robotaxi service in Austin is the closest real-world test of what unsupervised driving looks like. Launched in June 2025 with safety drivers, it began fully driverless rides in January 2026 and now covers 245 square miles of central Texas with approximately 20 vehicles.

The safety record tells a different story from the marketing numbers. Tesla reported 14 crashes to NHTSA across roughly 800,000 miles of Austin robotaxi operation through February 2026, a rate of one crash every 57,000 miles. That is approximately four times worse than the human average Tesla cites on its own safety page.

The fleet gap

Tesla had 42 vehicles authorised for driverless operation in Texas as of late May 2026, compared with Waymo’s 577. The fleet has been shrinking rather than growing, with Tesla deferring aggressive expansion until FSD v15 is ready, a timeline Musk has placed in late 2026 or early 2027.

The claim that roughly eight million Teslas are “running robotaxi-derived software” is technically true only in the sense that FSD Supervised shares a codebase with the robotaxi system. Those eight million cars operate at Level 2 with mandatory human supervision. Half of them, the four million with Hardware 3, cannot physically run unsupervised FSD regardless of software updates.

The contract rewrite

Between 2016 and early 2024, Tesla sold FSD as “Full Self-Driving Capability” with no mention of “supervised.” In March 2024 the product was renamed “Full Self-Driving (Supervised)” with version 12.3.3. In June 2026, owners discovered Tesla had retroactively modified old purchase agreements to insert the word “supervised,” making the original contract language inaccessible.

Tesla faces a certified class action lawsuit in the US over FSD false advertising and statements made between October 2016 and August 2024. The European Transport Safety Council has urged EU member states to take a “precautionary approach” to Tesla’s system, citing risks of driver over-reliance.

The flags

The 5.3 million miles-per-crash figure is self-reported by Tesla using a methodology that differs from NHTSA’s, and it describes FSD Supervised, not unsupervised driving. The only unsupervised data available, from the Austin robotaxi fleet, shows a crash rate approximately four times worse than the human average.

The claim that an owner drove 52,000 miles on FSD v14 without a single disengagement could not be independently verified. Documented records show a coast-to-coast run of 2,833 miles without intervention as the longest confirmed streak, and average intervention-free distances of around 25 miles.

Musk’s “acid test” of sleeping through your commute has been promised in various forms since 2014. The current timeline places unsupervised FSD for consumer cars no earlier than Q4 2026, and only for vehicles with HW4 hardware in geographically validated areas. Four million HW3 owners who paid up to $15,000 for FSD will need hardware upgrades that Tesla has not yet priced or scheduled.



Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


Global law enforcement operation takes First VPN offline

Pierluigi Paganini
May 21, 2026

Police seized First VPN in a global crackdown, exposed its cybercrime users, and shut down infrastructure tied to ransomware and data theft.

A major international law enforcement operation has taken First VPN offline, a service that had become a quiet staple for ransomware crews, data thieves, and other cybercriminals trying to hide in plain sight.

“The coordinated action took place between 19 and 20 May and targeted the infrastructure behind one of the most widely used VPN services in the cybercrime underground.” reads the press release published by Europol. “The gathered intelligence exposed thousands of users linked to the cybercrime ecosystem and generated operational leads connected to ransomware attacks, fraud schemes, and other serious offences worldwide.”

Authorities seized dozens of servers across 27 countries, arrested the administrator, and carried out a search in Ukraine, cutting off an infrastructure that had been used in a wide range of serious investigations.

The service marketed itself as a privacy-first VPN with no logging and no cooperation with law enforcement, which made it appealing not just to ordinary users but also to threat actors looking to mask their activity. That’s the uncomfortable part of the VPN story: the same tools that help people protect privacy on public Wi-Fi or work securely from home are also useful for criminals who want to conceal their origin, route traffic through different regions, and make attribution harder.

“For years, the service, known as ‘First VPN’, was promoted on Russian-speaking cybercrime forums as a trusted tool for remaining beyond the reach of law enforcement. It offered users anonymous payments, hidden infrastructure, and services designed specifically for criminal use.” continues the press release. “‘First VPN’ had become deeply embedded in the cybercrime ecosystem, appearing in almost every major cybercrime investigation supported by Europol in recent years. Criminals used it to conceal their identities and infrastructure while carrying out ransomware attacks, large-scale fraud, data theft, and other serious offences.”

Europol said the service name kept resurfacing in major cybercrime cases, and Eurojust confirmed that investigators had been building the case for years through a joint effort led by French and Dutch authorities. 

What seems to have made this case especially valuable for investigators is that they didn’t just shut the service down, they also got inside its infrastructure before it disappeared. That likely gave them access to user records, connection data, and other evidence that can be used to map criminal activity back to real people and devices.

Authorities dismantled cybercrime infrastructure, including 33 servers and a service based in Ukraine, and seized domains linked to the operation: 1vpns.com, 1vpns.net, 1vpns.org, plus associated onion sites. They also notified users directly and shared information on hundreds of accounts with international partners, which suggests this may lead to follow-on investigations well beyond the VPN itself.

The bigger lesson is simple: privacy tools are not the problem, but criminal operators often rely on the same infrastructure normal users trust. Once that infrastructure is compromised, dismantled, or logged, the illusion of anonymity can disappear very quickly.

“The operation has already generated significant operational results at Europol’s level:

  • 21 Europol-supported investigations advanced through the intelligence obtained.”
  • 83 intelligence packages disseminated;
  • information linked to 506 users shared internationally;

“For years, cybercriminals saw this VPN service as a gateway to anonymity. They believed it would keep them beyond the reach of law enforcement. This operation proves them wrong. Taking it offline removes a critical layer of protection that criminals depended on to operate, communicate and evade law enforcement.” said Edvardas Šileris, Head of Europol’s European Cybercrime Centre

Follow me on Twitter: @securityaffairs and Facebook and Mastodon

Pierluigi Paganini

(SecurityAffairs – hacking, First VPN)







Source link