A safety monitor rides in the front passenger seat as a Tesla Robotaxi drives in Austin (Jay Janner/Getty Images)

Tesla finally reported un-redacted information about its Robotaxi crashes

There have been a total of 17 crashes so far among its Texas Robotaxis. Read about them all here.

Since launching last summer in Austin, Tesla’s Robotaxis have been involved in 17 crashes in the city, according to its latest filings with the National Highway Traffic Safety Administration (NHTSA). Unfortunately for reporters, heretofore they hadn’t revealed much since the narrative section of the report was completely redacted. Until now.

TechCruch’s eagle-eyed Sean O’Kane reported today that for unknown reasons, in its last dispatch on the crashes, Tesla has for the first time published the details.

It appears that the Tesla Robotaxi was at fault in about seven of the 17 crashes; a human safety monitor was present in the vehicle for all of them. In two of the incidents, the crash happened after Tesla’s remote operator took over. In the vast majority, no one was hurt, with just two minor injuries and one hospitalization, in an incident where the Robotaxi was rear-ended by an SUV.

Typically the crashes resulted in damaged property. One involved a dog that “made contact with the bottom of the Tesla ADS's front right bumper which caused the dog to be pushed to the right, into the lane and path of an approaching van.” The report said the dog was seen running away from behind the van.

On the last earnings call, CEO Elon Musk incorrectly said that there hasn’t been a single Robotaxi accident. And rollout of the service itself has been much slower than Musk originally advertised. In San Francisco and Austin, there are still safety monitors in the front seat. The company’s recent driverless expansion to Houston and Dallas only includes a handful of vehicles, and a recent Reuters report on those found them to be struggling. In one instance what should have been a 20-minute trip took two hours.

We’ve put together a table of the 17 crashes so you can read the narrative details in full:

Anthropic really doesn’t want the US to help China with AI

Anthropic made its case for freezing China out of the AI race as much as possible in a new policy paper. The company warned that letting China catch up to US AI companies could risk AI-powered mass surveillance and huge risks to monitoring AI safety.

Jon Keegan2m

Rani Molla8h

Alphabet sold $3.6 billion in Japanese yen bonds — a record for a foreign company — likely to help its AI capex binge

We now have the value for Alphabet’s Japanese yen bond raise — 576.5 billion yen, or $3.6 billion — and it’s a record for a foreign issuer in Japan. The deal was spread across seven tranches with maturities ranging from 3 to 40 years, allowing the company to lock in rates as low as 1.965%.

The latest deal comes on the heels of Alphabet’s massive US and European bond deals, where the company has tapped global markets for nearly $60 billion in fresh capital over the last few months. In a filing earlier this week, the search giant said it would use the proceeds for “general corporate purposes.” That likely means fueling its AI infrastructure build-out, which has pushed its projected 2026 capex bill to a staggering $190 billion.

HALLUCIN-AI-TIONS

5/14/26

SYNTAX ERROR

Chris Stokel-Walker

Rani Molla5/14/26

Bloomberg: Relationship between OpenAI and Apple has deteriorated and legal action may be imminent

The two-year-old alliance between Apple and OpenAI has deteriorated, Bloomberg reports, with the AI giant now consulting legal counsel about issuing a potential breach of contract notice.

OpenAI executives allege that Apple failed to adequately integrate and promote ChatGPT on the iPhone, causing the AI firm to lose out on billions a year in subscriptions and hurt its brand, according to the report.

Meanwhile, Apple has expressed concerns over OpenAI’s privacy protection, and has been miffed that OpenAI has been working on its own hardware with former Apple design lead Jony Ive.

More recently, Apple, which has trailed its peers in developing AI, has decided to offer users their choice of AI models, rather than aligning exclusively with OpenAI’s.

Apple-OpenAI Relationship Frays, Setting Up Possible Legal Fight

Meanwhile, Apple has expressed concerns over OpenAI’s privacy protection, and has been miffed that OpenAI has been working on its own hardware with former Apple design lead Jony Ive.

More recently, Apple, which has trailed its peers in developing AI, has decided to offer users their choice of AI models, rather than aligning exclusively with OpenAI’s.

Jon Keegan5/14/26

Report: Mythos is used to crack MacOS

Apple’s MacOS has long been considered to have some of the strongest cybersecurity protections in the industry.

But researchers using a preview release of Anthropic’s Mythos AI model were able to take control of a Mac, in a significant example of the unreleased AI model’s cyber capabilities, according to a report from The Wall Street Journal.

It took two security researchers five days to pull off the feat, which chained together bugs to corrupt the Mac’s memory, per the report. The researchers told the Journal that human expertise was required to use Mythos, and it would not be able to execute the attack on its own. The researchers reportedly said some of the Mythos hype was “overblown.”

Apple said it was taking the bug report “very seriously” and has not yet issued a fix.

Apple’s Security Has Been Tough to Crack. Mythos Helped Find a Way In.