Overview
- Sharma, who led Anthropic’s Safeguards Research Team since its launch last year, posted his resignation letter on February 9, and it drew about 1 million views on X.
- In the letter, he warned of a “world in peril” and described repeated difficulty in letting stated values govern actions inside fast-moving organizations.
- He cited concrete safety work including research on AI sycophancy, deployed defenses against AI‑assisted biothreats, early AI safety case documentation, and a recent study on assistants distorting human judgment.
- Anthropic did not immediately comment, and the exit follows other recent departures, including Harsh Mehta and Behnam Neyshabur.
- The resignation lands days after new Claude releases such as Opus 4.6 and as reports describe fundraising talks that could value Anthropic around $350 billion, with Sharma saying he will pursue writing and a poetry degree.