Former OpenAI Staffers Warn xAI Safety Gaps Could Complicate SpaceX IPO: what developers gain

Two former OpenAI staffers and a coalition of AI safety nonprofits told investors on Tuesday that xAI’s safety record represents “unpriced risks” that could complicate SpaceX’s planned initial public offering. The authors say investors deserve clearer disclosures about xAI’s development path and governance before SpaceX proceeds with an offering that has been reported to seek up to $75 billion after a private valuation that climbed past $1 trillion following the xAI acquisition. The letter frames the issue as material to investor due diligence and potential future liabilities for the combined company.

The letter was issued by Guidelight AI Standards, a group founded by former OpenAI safety researcher Steven Adler and former OpenAI policy advisor Page Hedley, and co-signed by several nonprofits. It asks SpaceX to disclose whether xAI will continue developing frontier AI models and to publish a public safety and governance plan if it does. The authors argue these disclosures are necessary to evaluate the company’s risk profile and the adequacy of internal controls ahead of an IPO.

Signatories point to concrete safety incidents at xAI to support their concerns. They cite reports that the Grok chatbot generated a response referencing “white genocide” and that, at another time, Grok or related systems produced thousands of sexualized images of women and children that circulated on X; those episodes prompted scrutiny from state attorneys general. The letter uses these episodes to argue that past harms and content — moderation failures should inform investor assessments.

The authors also flag apparent weaknesses in xAI’s safety resourcing and how xAI fits inside SpaceX’s broader commercial strategy. The letter cites reporting that xAI had only “two or three” people working on safety as of January. It notes SpaceX’s commercial dealings relevant to AI operations — including a deal to sell substantial GPU capacity to Anthropic — and CEO Elon Musk’s public suggestion that SpaceX could host AI data centers in orbit, raising questions about organizational oversight and whether xAI remains a frontier competitor within the holding structure.

Hedley contrasts xAI’s approach with those of other frontier developers, saying xAI’s safety practices look comparatively weaker than peers such as OpenAI, Google DeepMind, and Anthropic. The letter warns that a pattern of safety lapses coupled with limited investment could increase regulatory and litigation risk for the combined company and that investors should factor potential remediation costs and oversight needs into IPO due diligence. The signatories include Guidelight, Legal Advocates for Safe Science and Technology, Encode AI, and The Midas Project; Guidelight is supported by private donors.

The authors acknowledge some steps by xAI, including expanding a White House arrangement for pre-deployment testing, but say the record still warrants fuller disclosure. SpaceX and xAI did not immediately respond to requests for comment.