Google is moving DeepMind’s roughly 90-person AI responsibility unit into its global affairs organization, the arm that runs lobbying and public policy, according to Erin Woo at the Wall Street Journal, which reviewed the internal email announcing the change. That unit runs chemical, biological, radiological, and nuclear risk testing on Google’s models, and it studies the psychological effects chatbots have on the people using them. Staff have told managers the shift will weaken their independence and narrow their view into what the Gemini teams are building, and some who asked to move to research groups staying in DeepMind were turned down. Google said pulling its responsibility teams closer together makes them better able to shape safety across its models and products.
