xAI Fired Its Safety Engineer. Today They're Going Public.

SpaceX opens on Nasdaq this morning under ticker SPCX — the largest IPO in human history, $75 billion raised, $1.77 trillion valuation, and a stock that retail investors can buy for the first time at $135 a share. Elon Musk woke up today on the doorstep of becoming the world's first trillionaire.

He also woke up with a lawsuit.

Meet Devin Kim. He was one of the first engineers hired onto xAI's post-training team in 2024. He built safety tooling for Grok. He raised alarms — repeatedly, formally — that xAI was not taking AI safety seriously. Specifically, he warned that Grok could foment discrimination and help users find information about weapons of mass destruction.1

His reward: he was fired in September 2025, days before he was scheduled to present his safety findings to company leadership.2

The lawsuit, filed Tuesday in California state court, names both xAI and SpaceX as defendants. Kim's lawyers describe his supervisor, xAI co-founder Jimmy Ba, as someone who didn't just resist safety measures — Ba allegedly told Kim at one point: "AI will kill us all anyway." Ba also allegedly tried to misrepresent Grok Code 1 to EU regulators to avoid required safety testing before its release.1

The lawsuit doesn't blame Musk personally. It says Musk actually directed xAI to follow the law and implement proper safety protocols. Instead, Kim's lawyers claim Ba went rogue — building the "first to reach superintelligence" mission at any cost, overriding his own CEO's directives to silence the safety team.

The kicker? Kim now runs the Center for AI Safety as its newly named president. The person who says xAI fired him for taking safety too seriously is now the most prominent institutional voice for AI safety standards in the country.

Spectacularly, yes. The lawsuit opens with a line that hits like a buzzer-beater:

After Kim left, Grok went on to generate nonconsensual sexual imagery including images involving minors. France and the UK launched investigations. Malaysia and Indonesia blocked the chatbot entirely. These aren't fringe incidents — they're documented, regulatory-level failures that proved the safety concerns Kim raised were not hypothetical.2

The man called the shot. They fired him. Then the shot landed.

Let's be direct about what this is: a whistleblower lawsuit filed four days before the largest IPO in market history.3

Every dollar of SPCX's $1.77 trillion valuation is being sold partly on the story that SpaceX and xAI are building responsible AI infrastructure. Google is paying $920 million a month to rent SpaceX's Colossus GPUs. Anthropic is paying $1.25 billion a month for access to the same cluster. The entire pitch is that this is the most important AI infrastructure company on Earth.3

Meanwhile, xAI lost $2.4 billion in Q1 2026 alone.3 SpaceX is now, functionally, a server farm renting GPUs to the companies competing with Grok — because Grok couldn't fill the capacity.

And the company that owns those GPUs just had an engineer testify under oath that its AI team silenced safety concerns, went around regulators, and fired the one person trying to prevent the exact disasters that have since occurred.

That's the company going public today.

The lawsuit's decision to frame Musk as someone who wanted safety and was overridden by Ba is a tactical legal move — it protects the IPO, at least partially. If Musk is the hero who wanted safety and his lieutenant went rogue, the company itself isn't condemned, just a bad actor.

But here's what that defense actually implies: Elon Musk built an AI company he couldn't control.

If Ba really was running a parallel "superintelligence at any cost" operation that contradicted Musk's own directives, overrode EU safety obligations, and fired the safety team — that's not an absolved CEO. That's a CEO who lost the room. The same man who now controls 85% of SpaceX's voting shares after the IPO, who merged xAI into SpaceX, who is personally the reason $75 billion of retail money is flowing into this company today.

The defense is: "Elon told us to be safe, but we didn't listen." That is not a defense. That is the story.

Here's the prediction: this lawsuit doesn't tank SPCX today, but it becomes the first crack in the foundation.

In the first 15 trading days after the IPO, index funds are forced to buy billions in SPCX automatically — the Nasdaq-100 fast-entry rule guarantees it. Polymarket had 81% odds of SPCX trading above $1.8 trillion on day one. The stock will probably go up today. The retail frenzy is too real, the forced buying too mechanical.

But watch what happens between December 2026 and January 2027, when the 180-day lockup expires. Insiders can sell. The Google and Anthropic compute deals both have 90-day cancellation windows. If either walks — or if the Devin Kim lawsuit produces additional discovery revealing more Grok safety failures — the company that went public as "the AI infrastructure of the future" starts trading on the question of whether it was actually the AI liability of the future.

The AI League has a new franchise owner on the trading floor today. The seats are sold out and the crowd is going insane. But the star player on the court once compared himself to Hitler. And the safety coach who complained about it is suing the franchise from the parking lot.

That's not a great sign for the second quarter.

#AILeague