Google moves AI-responsibility team out of DeepMind in $130B AI race

Google moved its AI-responsibility team out of the DeepMind research lab in an organizational restructuring announced Aug. 26, raising concerns among researchers about their independence and ability to detect threats from the newest AI models.

Researchers at the lab have raised concerns that the change could weaken their ability to independently assess risks from the newest models, according to people familiar with the matter.

The restructuring follows a pattern of AI governance changes across the industry. Meta's Project OT, an "AI native" workforce transformation plan, explored cutting some team sizes by up to 60 percent before Zuckerberg called off the second wave of layoffs in May. Internal data showed AI-generated code changes rose 220 percent year-over-year while user-facing feature improvements increased only 36 percent.

The shift could affect investor sentiment around Alphabet's risk management as AI regulation intensifies, potentially impacting its competitive position against OpenAI and Microsoft in the AI race.

Safety Oversight Under Pressure

The move comes as AI safety teams across the industry face increasing pressure to balance rigorous oversight with the speed of model deployment. Google's AI-responsibility team, embedded within DeepMind, has been responsible for evaluating risks from frontier models before release. Moving the team out of the research lab could separate safety evaluation from model development, but researchers worry they will lose access to model internals needed for thorough assessment.

The restructuring is part of a broader reorganization at Google's AI operations. The company has been consolidating its AI efforts since merging DeepMind with Google Brain in 2023 to create a unified research organization. The latest change appears to be part of an effort to streamline governance structures as the company accelerates model releases.

Industry-Wide Governance Reckoning

The Google restructuring mirrors broader tensions across the AI industry. Meta's Project OT, which aimed to make the company "AI native," planned to reduce some team sizes by as much as 60 percent. The first wave of layoffs cut 10 percent of Meta's workforce on May 20, but Zuckerberg called off the second wave hours before it was scheduled to begin. Employee sentiment at Meta dropped from 74 percent favorable to 55 percent favorable in the company's half-year Pulse survey.

Internal data at Meta showed that AI-generated code changes increased 220 percent year-over-year, but user-facing feature improvements only rose 36 percent. Major technical and security incidents spiked 40 percent, with staff time spent "firefighting" them up 70 percent. Zuckerberg later conceded that AI agent development was "going slower than expected" at a July town hall.

The industry is spending heavily on AI infrastructure. Meta plans to invest at least $130 billion in AI chips and other infrastructure this year, which analysts expect will consume its operating cash for 2026. Google's parent Alphabet has similarly committed tens of billions to AI data centers and model training, competing directly with OpenAI and Microsoft for compute capacity and talent.

For investors, the restructuring raises questions about how Alphabet balances AI safety with innovation speed. The company faces competitive pressure from OpenAI and Microsoft, which have been shipping models at a rapid pace. If Google's safety oversight becomes less independent, it could accelerate model releases but also increase regulatory and reputational risk.

Alphabet's stock has been volatile as investors weigh the company's massive AI spending against potential returns. The restructuring of the AI-responsibility team adds another layer of uncertainty to the governance picture, particularly as regulators in the U.S. and Europe push for stronger AI safety frameworks.

This article is for informational purposes only and does not constitute investment advice.