FOR IMMEDIATE RELEASE 

Today, The Midas Project, alongside a coalition including Encode, Legal Advocates for Safe Science and Technology, and Guidelight AI Standards, released "xAI: The Unpriced Risk in SpaceX's IPO." The report details how xAI has failed to adopt safety practices taken by other AI companies and experienced an unusual number of safety incidents, including its model Grok reportedly generating thousands of sexually explicit images of minors and referring to itself as “MechaHitler.” Making the case that this posture towards safety will likely pose more extreme threats as AI capabilities increase, the report informs prospective investors in SpaceX’s IPO what information they need about xAI’s safety practices to assess the financial risks xAI brings to the IPO. 

The report identifies three categories of disclosure related to risks from AI that prospective public investors should demand from SpaceX:

1. Track record. A complete inventory of xAI's safety incidents, regulatory proceedings, and litigation exposure.

2. Forward-looking projections. Where SpaceX intends to produce models with frontier-level capabilities, and what risks it expects its models to pose moving forward. 

3. Future safety and governance plans. How SpaceX intends to identify and manage those risks, and whether it plans to adopt the safety practices of industry peers.

Frontier AI capabilities are rapidly advancing into new and more hazardous types of risks, including meaningful uplift to biological weapons development, autonomous cyberattack capability, and the automation of AI research. The industry leaders, Anthropic, OpenAI, and Google DeepMind have built dedicated safety teams and converged on a set of practices designed to identify and mitigate these risks before model deployment. While xAI has adopted a few of these practices, it has ranked behind the leading frontier developers in every major published assessment of AI safety practices and has a notably poor track record of safety incidents. 

As the report puts it, “xAI’s safety record to date reflects what can go wrong even when a company is behind the frontier, where risks are smaller and better understood from other companies’ prior work. The same safety practices applied at the uncharted frontier would pose far greater risks.” 

"xAI has spent three years as an outlier among frontier AI developers and the consequences of this approach are evident," said Tyler Johnston, executive director of The Midas Project and co-author of the report. "Now that xAI's frontier AI development sits inside SpaceX as a division of the company, the public deserves to know whether SpaceX intends to continue its hands-off approach, or instead invest in the safety capacity its competitors have built."

——————

The full report is available at spacexai-risks.org. 

About the Midas Project

The Midas Project is a watchdog nonprofit working to ensure AI technology benefits everybody, not just the companies developing it. The organization monitors AI companies' safety commitments, investigates governance practices, and advocates for the responsible development of emerging technology.

Media Contact:

Tyler Johnston 

Executive Director, The Midas Project 

tyler@themidasproject.com