Sharma, who built defenses against AI-assisted bioterrorism and pushed for transparency on model risks at the San Francisco AI firm, announced his resignation on Monday. He described struggles to let values guide actions amid mounting pressures, planning to return to the UK for a poetry degree and step back from the spotlight. His exit follows other safety team departures amid Anthropic's launch of Claude Opus 4.6 and a massive $20 billion funding round at $350, $350 billion valuation, fueling debates on balancing safety with commercial speed.