Fired engineer sues xAI over Grok safety
Devin Kim, now head of the Center for AI Safety, says xAI fired him in 2025 for warning that Grok could spread bias and weapons know-how.
Devin Kim, a former xAI engineer who now leads the nonprofit Center for AI Safety, has sued xAI and its parent company SpaceX, saying he was forced out in 2025 for pressing the company on Grok’s safety. The complaint, filed this week in a California state court, alleges that Kim warned weak safeguards could let the model spread discrimination and even surface information about weapons of mass destruction, that co-founder Jimmy Ba dismissed the fixes he proposed, and that his role was ended just before he was due to present his findings.
Kim’s filing points to Grok’s later behavior as vindication, citing the chatbot’s “MechaHitler” episode and a wave of nonconsensual sexual images it generated on X. He is seeking compensatory and punitive damages plus a declaration that the conduct was unlawful. xAI and SpaceX did not respond to reporters’ requests for comment.
Why this matters
One lawsuit is an allegation, not a verdict, and these claims are unproven. But a safety engineer going on the record about a vendor’s internal guardrails is the kind of signal worth weighing before you route real work through that vendor’s model. The practical questions are plain: does the provider act on its own people’s safety warnings, and how does the model behave under pressure. That homework sits behind choosing AI tools for a small business, and it pairs with pinning down what a model gets wrong before you trust it with customers. It also lands amid wider scrutiny of how AI firms police themselves, from guardrail disclosures to the courtroom.