Neil Chowdhury announced in a personal blog post that he has joined Thinking Machines Lab as a Member of Technical Staff, working on AI safety and alignment.
Chowdhury said he will focus on safeguards that can work when powerful models are open and customizable, rather than relying only on a developer to monitor usage and restrict access. He argues that solving those problems could let more people shape and inspect the AI systems they use without giving up protection against dangerous misuse.
Before the move, Chowdhury was on the founding team at Transluce, where he researched ways to automatically uncover unexpected and harmful model behavior. He previously worked on safety research at OpenAI, including SWE-bench Verified, a human-checked test of AI coding ability.