Anthropic's AI safety lead resigned Monday with a public warning that "the world is in peril" from interconnected global crises. Mrinank Sharma, who led the Claude chatbot maker's Safeguards Research Team, posted his two-page resignation letter on X where it was viewed over a million times.
Sharma's final day at the San Francisco-based AI company was February 9, according to his public announcement. The Oxford-educated researcher joined Anthropic in August 2023 after completing a PhD in machine learning at the University of Oxford.
"The world is in peril," Sharma wrote in his resignation letter. "And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment."
He warned that humanity appears to be approaching a threshold where wisdom must grow alongside technological capacity.
During his tenure, Sharma worked on defenses against AI-assisted bioterrorism, studied AI sycophancy where chatbots overly flatter users, and researched how intensive chatbot use can distort human perception of reality. His final project examined how AI assistants could "make us less human or distort our humanity."
Sharma cited a disconnect between stated values and actual practices.
"Throughout my time here, I've repeatedly seen how hard it is to truly let our values govern our actions," he wrote. "I've seen this within myself, within the organization, where we constantly face pressures to set aside what matters most."
The resignation comes as Anthropic transitions from a safety-focused research lab to a commercial powerhouse reportedly seeking a $350 billion valuation. The company recently launched Claude Opus 4.6, an upgraded model designed for high-end coding and workplace productivity. Anthropic recently revealed that its AI now writes nearly all of its internal product code, boosting productivity and enabling advanced security reviews.
Sharma plans to return to the UK and pursue poetry, writing, and "courageous speech." He mentioned studying poets like Rilke and William Stafford, whose poem "The Way It Is" he included in his resignation letter. The researcher said he wants to explore "poetic truth alongside scientific truth" as frameworks for understanding technological challenges.
His departure follows other high-profile exits from Anthropic. R&D engineer Harsh Mehta and AI scientist Behnam Neyshabur left last week to "start something new," while former AI safety researcher Dylan Scandinaro joined OpenAI as head of preparedness.
Anthropic has positioned itself as a leader in responsible AI development, frequently highlighting safety commitments. The company recently announced a fellows program offering $2,100 weekly stipends to researchers working on AI safety projects. Applications for the May 2026 cohort closed, but the program continues accepting applications for July 2026.
Sharma's resignation letter, complete with footnotes and poetry references, sparked widespread discussion about AI safety culture. Some observers noted that "AI safety resignation letters" have become a distinct genre, following similar departures from OpenAI and other labs.
The researcher's study published last week revealed that using AI chatbots could cause users to form distorted perceptions of reality, with thousands of such interactions occurring daily. His findings highlighted the need for AI systems designed to "robustly support human autonomy and flourishing."
Anthropic, founded by former OpenAI researchers, develops the Claude AI chatbot family. The company has partnerships with Amazon and Google and recently signed agreements with the U.S. AI Safety Institute for collaborative research on AI safety testing and evaluation. The importance of AI security was further highlighted by recent findings of hundreds of malicious extensions in the OpenClaw AI marketplace, posing severe data theft risks to users.
