Google moves AI-risk unit out of DeepMind, sparking internal concerns: WSJ

Google is reportedly moving the team responsible for risks and societal impacts of its AI development out of DeepMind, raising internal concerns.

According to the Wall Street Journal, the 90-person "AI-responsibility" team will move to Google's global affairs organization, which oversees the company's lobbying and public policy. Changes will take effect at the beginning of September.

This move has reportedly prompted members of the responsibility team to raise concerns that it will impede their ability to conduct independent research on the firm's AI development. The team assesses Google’s AI models for chemical, biological, nuclear, and radiological risks, alongside research on the psychological impacts of user interactions with AI.

The report said the global affairs organization includes Google's central trust and safety team. Other AI safety teams, including ones for model behavior, privacy and security, continue to operate under DeepMind.

Earlier this month, DeepMind underwent a restructuring that had co-founder and CEO Demis Hassabis move into a new chairman role. Koray Kavukcuoglu will succeed Hassabis as senior vice president, rather than CEO.

Google DeepMind Vice President Helen King told the team that the changes will not impact day-to-day operations and mission, WSJ reported. King added that Hassabis remains committed to responsible AI and plans to stay involved.

Google acquired DeepMind in 2014 and allowed the organization to operate in a semiautonomous manner. This restructuring is seen as a move to have DeepMind more closely integrated with the rest of Google, according to the report.

DeepMind has seen a string of top talent departures in recent months as its flagship AI models fell behind competitors like OpenAI and Anthropic in key capabilities.

The wave of departures includes veteran Google executive Jeff Dean, who exited earlier this month alongside three senior researchers to launch a startup. His exit follows the high-profile departures of prominent researchers Noam Shazeer to OpenAI and John Jumper to Anthropic earlier this summer.

Alphabet (GOOG) fell 1.23% on Wednesday to close at $339.10.