Google Moves AI Responsibility Team Out of DeepMind, Raising Safety Independence Concerns

Google is transferring its 90-member "AI Responsibility" team from DeepMind to its central Global Affairs department, which manages lobbying and public policy. This organizational shift coincides with a leadership change where DeepMind co-founder Demis Hassabis steps down as CEO to become Chairman. The reorganization sparks internal concerns that moving safety researchers to a policy-focused corporate department will compromise the independence of AI safety evaluations. It could also restrict researchers' direct access to core model development teams, potentially hindering their ability to identify and mitigate risks in models like Gemini. While the AI Responsibility team will retain access to DeepMind's computing resources, other safety-focused teams covering model behavior, privacy, and security will remain within DeepMind. The transition has already prompted fears of talent attrition, following the recent departures of several high-profile AI researchers from Google.