Been Kim says her team at Google DeepMind “completely pivoted to safety” about two months before her November 19, 2026, post—and she thinks others should follow. She also announced full-time, Seattle-only hiring to tackle urgent safety problems directly on Gemini. She's asking for reinforcements. Source
Google's linked Careers listing identifies Kim as the principal scientist leading the Deep Understanding team. Kim described a pivot within her team, not a Google DeepMind-wide change of direction. SourceSource
“Interpretability is still dear to my heart, but it's always been a means to an end—and right now, I can't justify working on anything other than directly trying to improve frontier model safety,” Kim wrote. Source
The job listing says: “We prioritize concrete safety wins over conventional paper-writing cycles.” It describes day-to-day work on “real vulnerabilities in live systems,” alongside mid- and long-term safety questions for future frontier models. Source
A paper can still be on the agenda; it isn't the finish line. The same listing says researchers will share and publish findings, and its responsibilities include authoring research papers. Source
The listing accepts research experience in AI safety, alignment or mechanistic interpretability, and names agentic or mechanistic interpretability for diagnosing model behavior among its preferred qualifications. Interpretability hasn't disappeared from the paperwork. Neither Kim's post nor the job listing specifies which methods the team is using now. SourceSource
Kim's post and the job listing don't identify a specific trigger for the pivot, the team's first project, particular Gemini vulnerabilities or measured safety outcomes. SourceSource
Kim directed applicants to the linked role and said she wouldn't be able to reply to emails. The appeal comes with an application route, not an invitation to fill her inbox.
Comments