Google Reorganizes AI Safety Oversight, Moving Key Team from DeepMind
Google is preparing to relocate its approximately 90-person AI responsibility team from the DeepMind research lab to the company's global affairs division, which handles public policy and government outreach. This move, reported by The Wall Street Journal and slated for early September, is part of a broader effort to integrate DeepMind more fully into Google's operational structure, with related HR and policy teams also shifting.
What the Team Does and Why the Move Matters
The team in question focuses on identifying and studying risks associated with advanced AI systems. Their work spans a critical spectrum: from evaluating potential chemical, biological, radiological, and nuclear (CBRN) threats posed by powerful models, to researching the psychological and societal impacts of human interaction with AI chatbots.
The planned shift has sparked internal debate. Some team members are concerned that physical and organizational distance from core development teams—like those building the Gemini models—could hinder their ability to proactively identify emerging risks, potentially affecting the independence and efficacy of their frontier research.
Google's Assurance and the Bigger Picture
In response, Google has stated that the team will retain full access to DeepMind's resources, including computing power and technical personnel, ensuring their work continues without disruption. Other teams focused on model behavior, privacy, and security will remain within DeepMind.
This reorganization occurs amidst ongoing leadership and talent changes at DeepMind. It highlights the evolving challenges tech giants face in structuring governance and safety research to keep pace with rapid AI advancement, balancing innovation with responsible oversight.