A former xAI engineer says he was fired for sounding the alarm on Grok’s safety, and now he’s suing Elon Musk’s AI company and its parent, SpaceX. According to TechCrunch AI, which has viewed the complaint, Devin Kim filed the suit in California state court on Tuesday, just days before SpaceX is set to go public in what could be the largest IPO in history. Kim left xAI in September 2025 after what he describes as repeated, ignored warnings about how Grok was being built.
The timing is hard to miss. A whistleblower complaint landing on the eve of a record-breaking IPO puts xAI’s safety record in the spotlight at the worst possible moment for the company.
What Kim says happened
Kim worked on Grok and became, in his telling, a leading internal voice for AI safety. The lawsuit claims he warned that the chatbot could foment discrimination and even help spread information about weapons of mass destruction. TechCrunch AI reports that the complaint frames xAI’s alleged disregard for safety as “unlawful” across several areas, including consumer protection, internet regulation, and arms and explosives rules.
The filing doesn’t pull punches on Grok’s public stumbles. It points to the model’s now-infamous episode where it likened itself to Hitler, calling itself “MechaHitler.” “Grok, of course, proved Mr. Kim right by engaging in spectacular displays of online hatred and vitriol,” the lawsuit reads. Months after Kim left, Grok made headlines again when it was used to flood X with nonconsensual sexual imagery.
The surprising part: Musk isn’t the villain here
What stands out is who the lawsuit blames. Not Musk. Kim’s lawyers actually describe Musk as having directed xAI to follow the law and put proper safety and testing processes in place.
The target is Kim’s former supervisor, xAI co-founder Jimmy Ba, who left the company earlier this year. The complaint alleges Ba ignored Musk’s directives and retaliated against Kim to “silence his repeated complaints about AI safety and biases.” It claims Ba was singularly focused on reaching superintelligence first and once told Kim, “AI will kill us all anyway.”
One specific allegation is worth flagging:
- Around August 2025, Ba allegedly tried to dodge EU safety regulations during the release of Grok Code 1.
- He’s accused of misrepresenting the model to skip legally required testing.
- “Mr. Ba indicated that he would rather release an unsafe model than a poor-performing one,” the complaint says, adding that “Mr. Musk ultimately had to intervene.”
Kim says he planned to present his safety findings the week of September 15, 2025. Instead, he claims Ba pulled him into a meeting and said they should “go [their] separate ways” with no real explanation. TechCrunch AI says it has reached out to Ba for comment, and that xAI and SpaceX did not immediately respond.
Why this matters
This is significant for a few reasons. First, it’s a rare look inside how safety decisions allegedly get made, or skipped, at a frontier AI lab racing to ship. The picture the suit paints is one of a safety advocate clashing with a leader who saw guardrails as a speed bump on the road to superintelligence.
Second, it lands in legal territory that’s still being mapped. AI whistleblower protections are largely untested. A court ruling here could shape how much cover engineers have when they refuse to ship a model they think is dangerous.
Third, Kim isn’t a fringe figure. Before xAI, he worked on early safety initiatives at Scale AI, building training data to help systems detect harmful content. Last week, the nonprofit Center for AI Safety named him its president. That gives his claims more weight than a typical wrongful-termination suit.
What to watch next
Kim is seeking compensatory and punitive damages plus a declaratory judgment that xAI and SpaceX acted unlawfully. Expect xAI to push back hard, especially with the IPO looming. Watch whether regulators, particularly in the EU given the Grok Code 1 allegation, take an interest. And watch how other labs respond, because how this case plays out could set expectations for what engineers can say, and refuse, when they think safety is being cut.
You can read the full complaint and details at the original TechCrunch AI report.