Google Moves Gemini's Bioweapon-Risk Safety Team Under Its Lobbying Chief, Denies Transfer Requests

Google is restructuring the team responsible for deciding whether Gemini could help someone build a bioweapon, and the new reporting line runs through the company's chief lobbyist.

Starting September 1, Google's roughly 90-person AI responsibility unit will leave Google DeepMind and report instead to Kent Walker, Google's President of Global Affairs, according to the Wall Street Journal, which first reported the move on August 27 citing an internal email. Walker's organization runs Google's relationships with regulators worldwide and shapes its lobbying and legislative strategy.

The unit in question evaluates Gemini for chemical, biological, radiological, and nuclear risks and separately studies how chatbots affect users psychologically. It's the team with authority to flag or halt releases over the kind of catastrophic risk that policy debates about AI regulation are built around.

Transfer requests denied

Several members of the team applied to transfer into research groups that will remain inside DeepMind before the September 1 restructuring took effect. Those requests were turned down, according to the Wall Street Journal's reporting.

Companies reorganize divisions constantly, and reorganization alone proves nothing. But when the people most affected by a change try to opt out and get told no, that's a specific, concrete signal about how the company views the move, regardless of what the press release says.

Helen King, the Google DeepMind vice president who leads the responsibility unit, told staff in an internal email that the team's research mission, computing resources, staffing levels, and ability to publish externally would remain intact. She described the change as part of DeepMind evolving toward a more conventional corporate structure, with support and policy functions centralizing under the broader Google organization.

A Google spokesperson gave the Wall Street Journal a brief statement saying the move was about "bringing our AI responsibility teams closer together." Google has not disputed the reporting on the denied transfers.

Why the independence question matters

Google's own chief scientist argued just six weeks before this reorganization that organizational independence was essential to responsible AI oversight. Moving the team's reporting line to the executive in charge of lobbying regulators is not a neutral administrative tweak.

A safety team that answers to research leadership and a safety team that answers to the lobbying shop are not the same thing, even if the researchers keep the same desks and the same computers. The concern critics will raise is straightforward: the unit that can flag Gemini for bioweapon risk now sits inside the organization whose job is to manage how governments regulate Gemini. King's email addressed continuity of mission and resources. It did not address that structural question directly.

No regulatory body has opened an inquiry into the restructuring, and nothing here alleges Google broke any law. This is a corporate organizational decision, not a legal violation, and there's no charge or finding to report because none exists.

The public doesn't trust any of this anyway

The timing lands against a backdrop of broad public skepticism toward AI companies generally. A Pew Research study found 52% of Americans are now more concerned than excited about AI's growing role in daily life, up from 37% in 2021, according to Euronews. Among adults under 30, 55% say they're more concerned than excited, a first since Pew began tracking the question.

A CNBC Generation Lab poll of 1,088 respondents found majorities distrust every one of nine major AI executives to act responsibly, per Euronews. Alphabet CEO Sundar Pichai was distrusted by around 75% of respondents. Microsoft's Satya Nadella fared best of the group and was still distrusted by 65%.

A separate bet: let users decide what's true

While Google centralizes its AI safety oversight under corporate leadership, a small startup is betting the opposite instinct sells: pushing trust decisions down to individual users instead of a company algorithm.

Stephen Coates, Twitter's former head trademark counsel, launched Twitter.now this week through his startup Operation Bluebird, according to Business Insider. The platform uses an AI system called VERA to assign posts a "trust score," letting users filter their own feeds by threshold rather than have a central algorithm hide or ban content outright.

"We're going to use the phrase 'freedom of speech and not freedom of reach,'" Coates told Business Insider. Early access costs $20, with a $40 "Fighter" tier, according to Dealroom.

X sued Operation Bluebird in Delaware last December seeking to block the launch, arguing it owns the Twitter trademarks. Operation Bluebird argues X abandoned "Twitter" and "Tweet" after rebranding, according to Dealroom and Ground News. No ruling has been issued. X has said publicly that "Twitter never left and continues to be exclusively owned by X Corp." The Delaware case remains unresolved, and how a judge rules on the trademark question will determine whether Twitter.now can keep its name.