Former OpenAI Employees Warn xAI’s Harmful Security Documents Could Derail SpaceX’s IPO

Two former employees of OpenAI and the AI ​​nonprofit group warn that Elon Musk’s AI lab, xAIcan be a liability to prospective investors in SpaceX, which is preparing to post what is expected to be the most public offering in Wall Street History.

In a letter In an investor briefing published on Tuesday, former employees highlighted what they called “priceless risks” related to xAI that could disrupt the plans SpaceX announced to implement. $75 billion as part of its IPO. The rocket company’s private valuation rose to $1 trillion later got xAI last year. Musk said his rocket company could build a data center in its AI space, but the authors of the letter said xAI’s poor safety record could cloud investors’ view of the company when it plans to deliver. IPO prospectus file.

One of the letter’s signatories and co-authors is a new non-profit organization called Guidelight AI Standards, which was founded by a former security researcher at OpenAI. Steven Adler and former OpenAI policy advisor Page Hedley. The group, which is supported by private donors, aims to develop ways to protect the frontier AI industry. Other AI security nonprofits also signed on, including Legal Advocates for Safe Science and Technology, Encode AI, and The Midas Project.

Hedley told WIRED in an interview that he believes xAI has more secure practices “almost across the board” compared to other frontier AI developers, including OpenAI, Google DeepMind, and Anthropic. As a result, he says, SpaceX may face greater regulatory and legal risks than other AI labs.

The authors of the letter say SpaceX should make several disclosures to investors, including whether xAI intends to continue developing frontier AI models. SpaceX soon struck a deal to sell a large portion of its GPU power to Anthropicand the letter says the deal “does not inform whether xAI is still an AI competitor within the larger company.” If xAI continues to create AI models of the border, the authors say that it should be published a policy of security and public administration.

SpaceX and xAI did not immediately respond to WIRED’s request for comment.

The letter also cites examples of how xAI has not maintained corporate security measures, such as publishing detailed risk mitigation plans around its AI models used in cyber attacks. The authors also describe specific security issues with xAI that they say warrant further investigation. Among the most popular is xAI’s flagship AI chatbot, Grok, itself brought about the killing of the whites in his answers. At one point, xAI allowed Grok to produce thousands sexual images of women and childrenwhich went viral on the social network of Musk X. The result caused at least 37 US Attorneys General sending a letter demanding that Musk’s AI lab take action to protect women and children on its platform.

Hedley says the number of security incidents xAI has experienced and the attention it has received “is very different from the rest of the market.” As policymakers become more and more aware of the cyber capabilities of advanced AI models like Claude Mythos of the Anthropicnew safety regulations may be on the horizon. The Trump administration says it already is monitoring the sugar system which would give US intelligence agencies oversight of AI models.

“It takes a lot of money to manage the risk of (AI security), and it looks like xAI has already been invested here,” says Adler. The letter also echoes the Washington Post’s claim that xAI had “two or three” people who have been working on security since January. “The question that businesses have to ask themselves is if xAI is going to stay on the edge, how expensive is it going to be to manage these (threats) properly? If they don’t, what will be the consequences?”