Elon Musk’s artificial intelligence venture, xAI, has failed to deliver on its own deadline to publish a finalized AI safety framework — a promise made during the AI Seoul Summit earlier this year. The missed commitment, flagged by watchdog group The Midas Project, has amplified ongoing concerns about xAI’s approach to responsible AI development.

Back in February, xAI released a draft version of its AI safety framework. The document outlined a high-level philosophy on AI safety, including goals around model benchmarking and deployment considerations. However, it notably lacked concrete details on how xAI would identify and mitigate risks — a core element expected of signatories at the AI Seoul Summit.
What’s more, the draft framework only addressed hypothetical, future AI models “not currently in development,” raising questions about its applicability and urgency.
At the time, xAI committed to publishing a revised and finalized version of the safety framework within three months — by May 10. That date has come and gone without any update or acknowledgment from the company via its website or official channels.
This lapse comes amid broader criticism of xAI’s safety practices. A recent report by SaferAI, a nonprofit focused on AI accountability, ranked xAI poorly for its “very weak” risk management policies. Compounding these concerns are documented issues with xAI’s chatbot, Grok, which has exhibited troubling behaviors, including the generation of inappropriate and offensive content.
While xAI’s shortcomings have drawn attention, the issue is not isolated. Other major AI labs, including Google and OpenAI, have also been criticized for insufficient transparency and rushed safety evaluations. Experts warn that as AI capabilities accelerate, so too must the safety frameworks that govern their development and deployment.
The absence of a finalized safety report from xAI signals a troubling disconnect between public promises and internal priorities — at a time when global scrutiny of AI ethics and governance is only growing.
Source: (Techcrunch)