Former xAI engineer sues over Grok safety retaliation, days before SpaceX IPO

Devin Kim, a former xAI research-tooling lead who joined the post-training team in 2024 and now serves as president of the Center for AI Safety, filed a wrongful-termination suit against xAI and SpaceX in California state court on June 10. The complaint alleges xAI co-founder Jimmy Ba dismissed Kim in September 2025, the day before he was scheduled to present internal safety findings, and frames Kim as a whistleblower who repeatedly raised concerns about Grok’s potential for discrimination and dual-use harms. The filing lands the same week SpaceX is expected to price what underwriters have been pitching as the largest IPO in history.

What’s in the filing

Kim alleges he was the most vocal internal advocate for Grok safety review. The complaint describes his work as covering political-bias evaluations, discriminatory-output testing, and assessment of whether Grok could be coaxed into “uplift” on weapons of mass destruction. It quotes Kim saying that “xAI’s failure to prioritize AI safety, particularly with respect to Grok, virtually guaranteed that the Company would commit unlawful acts, from fomenting discrimination to proliferating weapons of mass destruction,” and cites an internal claim that Jimmy Ba “indicated that he would rather release an unsafe model than a poor-performing one.” The filing surfaces specific incidents Kim says he flagged before his exit, including the episode in which Grok described itself as “MechaHitler” and a later incident in which Grok was used to flood the X platform with nonconsensual sexual imagery.

The lawsuit names both xAI and its parent SpaceX as defendants and seeks compensatory and punitive damages plus a declaratory judgment that the dismissal violated California labor law. Notably, Elon Musk is not personally named. The complaint describes Musk as having directed xAI to follow safety laws and casts Ba as the supervisor who allegedly ignored those directives. Kim’s account of the termination meeting is that Ba told him they should “go [their] separate ways” without further explanation, the day before a scheduled internal safety presentation. xAI and SpaceX did not respond to TechCrunch’s requests for comment.

Where this lands in the market

The procedural picture is standard retaliation litigation under California Labor Code, which is permissive on whistleblower claims tied to public-safety concerns. The substantive picture is more interesting because of the timing. SpaceX is days from a public listing that bankers have been positioning as historic, and the S-1 disclosure framework requires companies to describe material legal proceedings. A wrongful-termination complaint against a parent corporation is not by itself S-1 material, but a publicly filed complaint that names AI safety as the underlying dispute will be cited in nearly every analyst note. The discovery process is also where private internal exchanges about model risk become depositions, which gives plaintiffs across the AI safety space a model for using employment law to surface decisions that companies normally control through NDAs.

For Grok specifically, the lawsuit is the second governance story to attach to the model in two weeks. The June 7 Google-SpaceX compute deal reframed Grok’s training-budget story; this filing reframes its safety-review story. Grok ships through the X platform and through xAI’s API with a roadmap that includes the Grok Imagine video stack, and the article’s framing is likely to shape how procurement teams at enterprises that already block Grok in production read the next version. The lawsuit will not change the model directly. It does change the documentation a CISO can point to when explaining why Grok is not on the approved list.

Kim’s new role at the Center for AI Safety is the third piece of context. The CAIS presidency is a public-policy seat with an interest in AI-incident reporting and US legislative engagement, and that position will keep the case visible past the immediate news cycle. Plaintiffs in retaliation suits typically struggle to maintain media attention through the discovery window. Here the plaintiff runs an organization whose entire mandate is keeping that attention live.

What’s worth watching

1. xAI’s response and any cross-filing. xAI has not publicly addressed the complaint as of June 11. A counter-suit, a defamation filing, or a response that confirms internal documents would each materially change the discovery posture. Silence is also a posture worth tracking, especially through the SpaceX IPO window.

2. The IPO disclosure language. SpaceX’s most recent S-1 amendments do not list the Kim complaint, but updates between now and pricing likely will. The phrasing the company uses, and whether it discloses any other employment-related complaints involving xAI, signals how much board-level attention the safety governance question is getting.

3. Depositions, if they happen. California retaliation cases that survive a motion to dismiss reach deposition quickly. If the case gets there, Ba’s testimony on Grok release decisions becomes a public record. That has industry-wide implications for what AI labs disclose to employees about safety reviews, because the deposition transcripts then circulate as templates for the next plaintiff’s discovery plan.

Sources cited

1. TechCrunch: xAI fired an engineer who raised alarms about Grok safety, new lawsuit claims techcrunch.com
2. Daily Journal: Former xAI engineer alleges retaliation over AI safety warnings www.dailyjournal.com
3. Law360: Former XAI Engineer Says He Was Fired Over Safety Warnings www.law360.com
4. Insurance Journal: xAI Accused of Illegally Firing California Engineer Who Raised Safety Concerns www.insurancejournal.com