When the people whose job is to make sure AI is safe keep quitting and warning us it isn’t, we should probably listen.
This month, senior employees at OpenAI, Anthropic, and xAI all headed for the exits - some with dramatic public letters, others with quiet departures that spoke louder than words. The common thread: they no longer believe their companies can be trusted to prioritize safety over growth.
”The World Is in Peril”
Mrinank Sharma didn’t mince words. The head of Anthropic’s safeguards research team announced his resignation in an open letter that landed like a bomb on X.
“The world is in peril,” he wrote. “And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment.”
Sharma joined Anthropic after his PhD, specifically to work on AI safety. He led the team responsible for defenses against AI-assisted bioterrorism, understanding AI sycophancy, and building internal transparency mechanisms. This wasn’t a disgruntled employee who never bought in - this was someone who dedicated his career to the mission.
His diagnosis was damning: “Throughout my time here, I’ve repeatedly seen how hard it is to truly let our values govern actions.” The pressure to ship faster, to compete harder, to grow bigger - it erodes even the most sincere commitments to safety.
He’s leaving to study poetry. That tells you something.
OpenAI Dissolves Its Safety Team
The same week Sharma resigned, OpenAI quietly disbanded its Mission Alignment team - a seven-person group created in 2024 to ensure that development of AGI stays true to OpenAI’s founding mission of benefiting humanity.
The team wasn’t just downsized. It was eliminated entirely, 16 months after being formed.
OpenAI has now dissolved multiple safety-focused teams in the past two years. The superalignment team led by Jan Leike? Gone, after Leike resigned citing safety concerns. The Mission Alignment team? Absorbed into other groups, its specific mandate erased.
The company that once published its charter promising to “prioritize safety over benefit” keeps removing the people responsible for that priority.
The Pattern Nobody Wants to See
Zoë Hitzig, a research scientist at OpenAI since 2024, broadcast her resignation in a New York Times essay. Her concern: OpenAI’s emerging advertising strategy, which she sees as fundamentally incompatible with safe, user-aligned AI.
She’s not alone. Nearly half of xAI’s founding team has departed. Multiple safety researchers across the industry have left citing similar frustrations: the gap between what companies promise on safety and what they actually prioritize when money is on the table.
“I’ve seen this within myself, within the organization, where we constantly face pressures to set aside what matters most, and throughout broader society too,” Sharma wrote.
The International AI Safety Report Agrees
These departures aren’t happening in a vacuum. The 2026 International AI Safety Report, authored by over 100 AI experts led by Turing Award winner Yoshua Bengio, concluded that “new AI capabilities are unpredictable, the functioning of models is insufficiently researched, and economic incentives on the part of AI companies hinder transparency.”
Economic incentives hinder transparency. That’s the polite way of saying: companies have every reason to downplay risks and rush products to market.
The report notes that theoretical risks from AI - cyberattacks, bioweapon development, manipulation - have now been demonstrated in practice. The warnings aren’t hypothetical anymore. They’re incident reports.
Why This Should Worry You
When software engineers quit Facebook over content moderation, or Google employees resign over military contracts, those are ethical disagreements about how technology is used. Important, but manageable.
This is different. These are the people specifically hired to identify and prevent catastrophic risks. And they’re leaving because they’ve concluded the task is impossible under current conditions.
Sharma described how hard it is “in practice for a company to let our values govern our actions when the money, the market, and the internal prestige all point toward shipping more capable models, faster.”
That’s not a fixable management problem. That’s a structural incentive problem. And it applies to every major AI lab.
What’s Being Done (And Why It’s Not Enough)
Companies respond to these departures with press releases about safety commitments. New teams are formed to replace the ones that left. Safety boards meet quarterly to review responsible deployment.
But the people who know best - the researchers who built the safeguards - keep leaving. They’re not being replaced with people of equal expertise and commitment. The institutional knowledge walks out the door, and the warnings get filed under “former employee concerns.”
MIT researchers have called for AI whistleblower protections. The Future Society has published frameworks for internal dissent channels. But until there are meaningful external accountability mechanisms, the choice for safety-conscious researchers remains: stay and watch your concerns be overridden, or leave and hope someone listens.
Most choose to leave. And they’re all saying the same thing on their way out.
Maybe we should listen.