Google is moving a roughly 90-person AI responsibility team out of Google DeepMind to join the global affairs organization, a shift spurred by DeepMind CEO Demis Hassabis stepping aside into a chairman role. The reorganization aims to integrate DeepMind more tightly with the rest of Google, positioning Koray Kavukcuoglu as the new senior vice president. Helen King, who leads the responsibility team, stated the change will strengthen their ability to inform model safety, though some employees have raised concerns about reduced independence and access to frontier AI research.

The unit focuses on testing models for chemical, biological, radiological, and nuclear risks, as well as psychological impacts of chatbots. While King assured staff that computing resources and headcount would be retained, and that Demis Hassabis remains involved, the reorganization is scheduled to take effect in September. Employees expressed worries that the move could hamper their ability to identify relevant research questions amid mounting scrutiny of AI impacts.

Other AI safety teams focusing on model behavior, privacy, and security will remain within DeepMind. The global affairs organization oversees central trust and safety, product areas, HR, and policy. King emphasized that sharing work externally through publications and events continues to be a priority.