Who Guards the Guardians of AI?

Google’s decision to move its AI Responsibility team from the research-focused DeepMind to its Global Affairs division suggests a fundamental shift in how the company defines safety—not as a core engineering problem, but as a matter of public perception. This relocation of the team responsible for ethical oversight forces a necessary conversation about whether safety can truly flourish when it reports to the same division that manages a company's public image and lobbying efforts.

An organizational reshuffle announced in an internal email about the move will see the roughly 90-person AI Responsibility team integrated not with engineers, but with policy experts and lobbyists. For years, the argument for embedding safety teams directly within research divisions like Google DeepMind was proximity. Being close to the work meant having the context, relationships, and technical access to identify risks before they scaled into public-facing products. Removing the team from that environment creates a gap between the people building the technology and the people tasked with worrying about its effects.

The new home for Google's AI Responsibility team is the Global Affairs division, a group primarily focused on navigating public policy, managing the company's reputation, and engaging with regulators. While this signals that Google sees AI ethics as a major policy issue, it also raises concerns among some employees that the team’s purpose could subtly change. Instead of serving as an internal check on research, its work might be shaped to support lobbying efforts or to burnish the company’s image as a responsible actor, regardless of the internal reality. The fear is that ethics becomes a gloss, not a governor.

Can Safety Thrive When It Reports to Public Relations?

This organizational shuffle is more than a line-item change on a flowchart; it exposes the central tension within corporate AI development. The incentives for a research lab are to push boundaries and build more powerful systems. The function of a responsibility team is to counsel caution, to ask difficult questions, and, if necessary, to halt a project. Placing the institutional voice of caution under the authority of the division responsible for public perception creates a structural conflict. The success of Global Affairs is measured by a smooth public narrative and favorable regulation, which may not always align with the friction and delays that genuine safety oversight can create.

The true measure of an ethics team is not the elegance of its published principles but its ability to influence decisions when profits and prestige are on the line. As these systems become more integrated into our lives, the question of who holds the developers accountable—and from where—becomes one of the most important of our time. It’s a question of whether the guardians are positioned next to the engine, with a hand on the emergency brake, or in the press room, tasked with explaining why the crash was unavoidable.