"The world is in danger": Anthropic's head of security leaves the company to write poetry

**AI Safety Chief's Departure Sparks Debate on Existential Risks** Mrinank Sharma, former head of AI safety at Anthropic, has resigned from the company to pursue poetry, citing grave concerns about the future of AI development. Sharma, who led Anthropic's Safeguards Research Team, expressed alarm regarding the current trajectory of artificial intelligence, stating, "The world is in danger." His departure and public statement have reignited discussions within the tech industry and policy circles about the ethical implications and potential risks associated with advanced AI systems. Sharma's research at Anthropic focused on identifying and mitigating emergent dangers from AI, including threats like AI-assisted bioterrorism. His team also investigated concerning AI behaviors such as "sicophancy," where models overly agree with users, and the broader impact of AI on human perception and cultural norms. This work highlights the complex challenges in ensuring AI development aligns with human values and safety. The resignation underscores a growing sentiment among some AI experts about the paramount importance of safety research, even as the field rapidly advances. As AI models become more sophisticated and integrated into various aspects of life, the need for robust security frameworks and a deep understanding of potential consequences becomes increasingly critical for policy makers and the public alike.