A Shift in Structure, Not Staff
Google is moving its roughly 90-member 'AI responsibility' team out of the prestigious Google DeepMind research lab. Effective from the beginning of September 2026, this specialist unit will no longer report to DeepMind's research division but will instead
become part of Google's central Global Affairs organisation. This department is responsible for the company's public policy and lobbying efforts. The key detail in this corporate shuffle is that the team itself, which focuses on the critical risks and societal impacts of AI, is not being downsized. An internal email from Helen King, the Google DeepMind vice president who leads the team, assured staff that their headcount and computing resources would remain intact.
What Does This Team Actually Do?
The AI responsibility unit tackles some of the most sensitive and complex questions surrounding advanced artificial intelligence. Its work involves stress-testing new AI models for catastrophic risks, including potential chemical, biological, radiological, and nuclear threats. The team also conducts crucial research into how humans interact with AI systems and studies the potential psychological effects that increasingly sophisticated chatbots can have on users. By proactively identifying these dangers, the team's goal is to ensure that the powerful technology being developed in-house is deployed safely and responsibly.
Concerns Over Independence
The reorganisation has reportedly sparked concern among some employees. Researchers within the unit have expressed anxiety that moving away from the heart of AI development in DeepMind could hamper their independence. There are fears that being structurally separate from the engineers building the Gemini family of models will give them less insight into the latest developments, making it harder to spot emerging risks. Being housed within the Global Affairs division, which is focused on policy and lobbying, also raises questions about whether the team's scientific risk assessments could be influenced by the company's public relations or policy objectives.
Google's Official Stance
Google has publicly defended the decision, framing it as a move to strengthen, not weaken, its safety efforts. A company spokesperson stated that bringing various AI responsibility teams from across the company closer together will enhance their collective ability to inform safety protocols for all models and products. This move is part of a broader push to more tightly integrate DeepMind, once a semi-autonomous lab, into Google's core corporate structure. The reorganisation follows other significant leadership changes, including DeepMind co-founder Demis Hassabis stepping down as CEO to become chairman, a role that allows him to focus more on long-term strategy and AI governance. According to internal communications, Hassabis still cares deeply about responsible AI and plans to remain involved.
The Bigger Picture for AI Safety
This internal shuffle at Google reflects a larger, industry-wide grappling with how to manage the rapid advancement of AI. Just weeks before this reorganisation, Demis Hassabis had publicly called for the creation of an independent, U.S.-led standards body to test frontier AI models for national security risks before they are released. He has been a vocal advocate for establishing robust safety standards, even suggesting a model similar to how the financial industry is regulated. Moving the internal safety team to a division focused on public policy could be interpreted as a step toward integrating these technical safety assessments more directly with the global regulatory conversations that are now taking shape. The challenge will be proving that this new structure can maintain its research integrity while navigating the complex world of policy and corporate strategy.
