Anthropic AI researcher resigns over safety concerns to pursue poetry

Mrinank Sharma, a leading AI safety researcher at US-based company Anthropic, has resigned from his position, citing concerns about global risks including artificial intelligence, bioweapons, and interconnected crises. In his resignation letter shared on X, Sharma emphasized that while he enjoyed his work, he felt the need to step away and pursue a poetry degree and writing in the UK, describing a desire to ‘become invisible.’ Sharma had led a team investigating AI safety issues such as why generative AI systems tend to flatter users, potential AI-assisted bioterrorism, and the societal impacts of AI assistants on human behavior. His departure coincides with another resignation in the AI sector: Zoe Hitzig, an OpenAI researcher, cited ethical concerns over ChatGPT’s deployment of advertisements and the potential psychosocial effects of AI on users. Anthropic, known for its Claude chatbot and public benefit mission, has faced scrutiny in the past, including a $1.5 billion settlement over claims of using authors’ work without permission for AI training. The company continues to compete with OpenAI in AI innovation while highlighting safety-oriented approaches, though critics note ongoing tensions between ethical principles and commercial pressures. The resignations highlight the ethical and safety dilemmas emerging in the rapidly growing AI industry.