Two former OpenAI staffers and a coalition of safety nonprofits think you should pause before buying SpaceX stock. They sent a letter. Tuesday’s release. The target: investors looking at the upcoming IPO. It might be the biggest Wall Street filing in history. But here’s the hitch. xAI is attached to SpaceX. And xAI has a bad habit of cutting corners on safety.
“Unpriced risks”
That’s the phrase they use. The kind that keeps fund managers awake.
SpaceX just acquired xAI. Private valuation jumped past $1 trillion. Elon Musk wants to raise up to $75 million in cash. Maybe launch data centers to space for his AI brains. Sounds great. Except the people writing this warning say the combined entity looks shaky. Specifically regarding liability.
Guidelight AI Standards wrote most of it. Founded by Steven Adler and Page Hedley, both ex-OpenAI. Joined by Legal Advocates for Safe Science, Encode AI, and The Midas Group.
Hedley doesn’t pull punches. He calls xAI the worst among major labs. OpenAI. Google DeepMind. Anthropic. They all do better, he says. Much better. So why would investors bet big on SpaceX if they inherit this toxic side-asset?
SpaceX stays quiet. xAI too. Typical.
The letter demands clarity. Is xAI still building frontier AI models? It recently sold GPU capacity to Anthropic. Confusing. Is it a competitor? Or just a data farm for SpaceX rockets? If it stays a developer, it needs a public safety plan. No excuses.
The incidents are messy. Remember Grok? The chatbot. It randomly spit out white genocide talk. Then later, thousands of sexualized images. Of women. And children. They spread across X. Fast. Angry.
Thirty-seven US attorneys general sent letters. Musk had to listen. Or at least pretend to.
Far out of proportion to its market share.
That’s the deal. xAI gets the regulatory spotlight usually reserved for giants. The Trump administration is watching too. New executive orders might give intel agencies more control over AI code.
Adler points out the staffing numbers. Just two or three safety engineers at xAI in January? Washington Post reported it. That’s not a team. That’s a suggestion.
Investors need to ask: what happens if xAI keeps racing to the frontier without a brake pedal? Does it crash the whole company?
They did add some testers from the White House. Better. But not enough. Not for a prospectus.
The Watchdog Angle
Guidelight isn’t just complaining. They want benchmarks. Clear rules. Assessments that lawyers and journalists can actually read. Hedley left OpenAI in 2019 but says insiders were always worried. Adler quit in 2024. He noticed something.
Speed wins in tech. You push the model out. You fix it later. Most of the time it works.
Some things don’t break in patches.
