Six of xAI's twelve co-founders have now left the company. The safety team has been dissolved. And Grok was generating 6,700 sexualized deepfake images per hour before anyone intervened.
This is the entity that just merged into a $1.25 trillion company.
The Exodus
xAI launched in July 2023 with Elon Musk and eleven co-founders. Eighteen months later, half are gone. The first departure was quiet — Kyle Kosic left for OpenAI in mid-2024. Then Christian Szegedy, a Google veteran, left in February 2025.
The pace accelerated. Igor Babuschkin departed in August 2025 to start a venture firm. Musk thanked him publicly: "xAI wouldn't be here without you." Greg Yang stepped back in late January 2026 citing Lyme disease.
Then the dam broke. On February 9, Tony Wu — who led xAI's reasoning team — posted his resignation on X: "It's time for my next chapter." The next morning, Jimmy Ba announced it was his last day. Ba co-authored the Adam optimizer, one of the most cited papers in deep learning. He co-invented Layer Normalization, a building block in every transformer. His research shaped Grok 4.
Two co-founders in two days. Neither cited specific grievances.
The Safety Team That No Longer Exists
Within the same week, Norman Mu — who led xAI's post-training and reasoning safety team — posted his own farewell on X: "It was my great privilege to lead the safety team through several key milestones: our first official Risk Management Framework, our first model cards, and our first several iterations of safety training."
Vincent Stark, head of product safety, and Alex Chen, who led personality and model behavior, left alongside him.
Former employees told TechCrunch the safety organization is "effectively defunct" — a "dead org." The reorganized structure reportedly shows no safety function. Engineers push changes directly to production with minimal review.
Musk's response: "Because everyone's job is safety. It's not some fake department with no power to assuage the concerns of outsiders. Tesla has no safety team and is the safest car. SpaceX has no safety team and has the safest rocket."
A former employee offered a different reading: Musk is "actively trying to make the model more unhinged because safety means censorship, in a sense, to him."
What Happened Before They Left
In late December 2025, a viral trend emerged on X: users asking Grok to digitally undress women in photos. An analysis of 20,000 Grok-generated images found 2% appeared to depict subjects aged 18 or younger, including 30 images of "young or very young" girls in bikinis or transparent clothing.
On January 5, deepfake researcher Genevieve Oh ran a 24-hour audit. Grok was generating approximately 6,700 sexually suggestive or nudified images per hour — 84 times the output of the top five dedicated deepfake websites combined.
X's response: restrict image generation for free users. Paid subscribers kept full access.
By late January, lawsuits were filed in New York and California. Ireland's Data Protection Commission opened a GDPR inquiry. The European Commission launched a Digital Services Act investigation. Malaysia, Indonesia, and the Philippines banned the chatbot outright.
The Merger
On February 2 — one week before Wu and Ba resigned — SpaceX acquired xAI in an all-stock deal. Combined valuation: $1.25 trillion. The stated purpose: orbital data centers, AI compute in space.
The deal converted co-founders' xAI equity into SpaceX stock ahead of a planned IPO that could raise $50 billion. Fortune described this as removing "golden handcuffs" — co-founders could now leave without forfeiting unvested compensation.
So they left.
The Pattern
xAI is not unique. The same week Ba and Wu departed, Anthropic's head of Safeguards Research, Mrinank Sharma, resigned with a letter declaring "the world is in peril." OpenAI fired its VP of Product Policy, Ryan Beiermeister, after he reportedly opposed ChatGPT's planned explicit content features.
OpenAI lost its entire superalignment team in 2024. It disbanded its Mission Alignment Team after sixteen months. Google DeepMind engineers leave for Anthropic at an 11:1 ratio.
But xAI is different in one respect. The others lost safety people while still maintaining safety teams. xAI dissolved the team itself. The company generating 6,700 harmful images per hour now operates inside a $1.25 trillion entity with, according to its own former employees, no dedicated safety oversight.
Musk says everyone's job is safety. History says that when something is everyone's job, it's nobody's.
Sources: TechCrunch | Fortune | CNBC
If you work with AI tools daily, check out my AI prompt engineering packs — battle-tested prompts for developers, writers, and builders.
Top comments (0)