News

AI Safety's Exodus: The People Who Built the Guardrails Are Walking Away

A wave of safety researchers and executives have quit OpenAI, Anthropic, and xAI in the span of a single week, warning that the industry is moving too fast and abandoning its own principles.

By Elena Marchetti • • 6 min read
AI Safety's Exodus: The People Who Built the Guardrails Are Walking Away

In the span of a single week, the three most prominent AI labs in the world have lost key safety researchers and executives. Some left quietly. Others lit the building on fire on the way out.

The departures paint a picture that should worry anyone paying attention: the people tasked with making sure AI systems don't go off the rails are concluding, one by one, that the rails were never the priority.

Anthropic: "The World Is in Peril"

Mrinank Sharma, who led Anthropic's Safeguards Research Team since its formation in early 2025 and had been at the company since 2023, announced his resignation on February 9 in a letter posted to X.

His words were not subtle.

"The world is in peril," Sharma wrote. "And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment."

Sharma's team worked on some of Anthropic's most sensitive safety problems - defenses against AI-assisted bioterrorism, research into AI sycophancy (the tendency of chatbots to tell users what they want to hear), and what he described as one of the first formal AI safety cases. His departure is not the loss of a peripheral figure. It is the loss of someone who was at the center of the company's safety apparatus.

The most pointed passage in his letter dealt with internal culture: "Throughout my time here, I've repeatedly seen how hard it is to truly let our values govern our actions. I've seen this within myself, within the organization, where we constantly face pressures to set aside what matters most."

He added: "We appear to be approaching a threshold where our wisdom must grow in equal measure to our capacity to affect the world, lest we face the consequences."

His plan? Move back to the UK, pursue a poetry degree, and "devote myself to the practice of courageous speech." Anthropic told CNN that Sharma was not the head of safety, nor was he in charge of broader safeguards at the company.

OpenAI: Two Departures, Two Different Alarms

OpenAI is dealing with a more complex situation. Two separate departures, one voluntary and one not, have raised overlapping but distinct concerns about the company's direction.

The Researcher Who Quit Over Ads

Zoe Hitzig, an OpenAI researcher for the past two years, announced her resignation in a New York Times guest essay titled "OpenAI Is Making the Mistakes Facebook Made. I Quit." Her concern is specific and structural: OpenAI's decision to introduce advertising in ChatGPT.

Hitzig argues that ChatGPT has "generated an archive of human candor that has no precedent." Users share their medical fears, relationship problems, career anxieties, and spiritual beliefs with the chatbot because they believe it has "no ulterior agenda." Building an advertising business on top of that data, she warns, creates "a potential for manipulating users in ways we don't have the tools to understand, let alone prevent."

She is not anti-advertising in principle. "I don't believe ads are immoral or unethical," she wrote. "AI is expensive to run, and ads can be a critical source of revenue." Her concern is about what comes next: "I believe the first iteration of ads will probably follow those principles. But I'm worried subsequent iterations won't."

The parallel to Facebook's trajectory is deliberate. Initial privacy commitments, followed by slow erosion, followed by a business model fundamentally at odds with user interests. Hitzig sees the same playbook unfolding.

The Executive Fired Over "Adult Mode"

Then there is Ryan Beiermeister, who served as Vice President of OpenAI's Product Policy Team - the group that develops rules and safeguards for how the company's products can be used.

According to reporting by TechCrunch and Futurism, Beiermeister was terminated after expressing concerns about OpenAI's planned "adult mode" feature, which would allow sexually explicit conversations in ChatGPT. OpenAI says she was fired for allegedly sexually discriminating against a male employee. Beiermeister denies the allegation, calling it "absolutely false."

OpenAI has stated that "her departure was not related to any issue she raised while working at the company." The timing, however, is difficult to ignore. Beiermeister had specifically raised concerns about inadequate guardrails to prevent child exploitation content and teen access to adult material.

Fidji Simo, OpenAI's CEO of Applications, has confirmed the adult mode feature is planned for launch during Q1 2026.

xAI: Half the Founding Team Is Gone

Meanwhile, Elon Musk's xAI is bleeding co-founders. Tony Wu and Jimmy Ba both announced their departures on February 10, bringing the total number of departed co-founders to six out of the original twelve-person founding team. Five of those departures happened in the last year alone.

Wu's farewell was diplomatic - "It's time for my next chapter" - but the pattern is unmistakable. Igor Babuschkin, Kyle Kosic, and Christian Szegedy have also left. Greg Yang announced last month he would step back to focus on his battle with Lyme disease.

When half your founding team walks away in under three years, the question is no longer about individual motivations. It is about organizational culture.

The Pattern

These departures are not happening in isolation. They are happening simultaneously, across competing organizations, in the same week. That matters.

The common thread is a growing gap between what AI labs say they value and how they actually behave. Anthropic was founded explicitly as a safety-focused alternative to OpenAI. OpenAI was founded as a nonprofit to benefit humanity. xAI was founded with promises of truth-seeking AI. All three are now losing the people who were supposed to hold them to those founding visions.

The industry is in a fierce competition for market share, with Anthropic fresh off a $30 billion funding round at a $380 billion valuation, OpenAI planning an IPO for Q4 2026, and xAI racing to keep up. The incentives are clear: ship faster, monetize harder, worry about safety later.

Sharma's warning deserves to be repeated: "We constantly face pressures to set aside what matters most."

The people who built the guardrails are telling us, plainly, that the guardrails are being dismantled. The question is whether anyone with the power to do something about it is listening.

Sources:

Tagged
AI Safety OpenAI Anthropic XAI Ethics
About the author Senior AI Editor & Investigative Journalist

Elena is a technology journalist with over eight years of experience covering artificial intelligence, machine learning, and the startup ecosystem.