Demis Hassabis walked into a G7 working lunch this week and did something you don't usually see from the head of a frontier AI lab. He asked world leaders to regulate him.
The Google DeepMind CEO used the lunch to pitch a US-led global AI watchdog, first reported by Axios and then The Verge. He described it as "a body that increases trust" — a technical standards organization with international membership but American leadership, capable of updating its rules as fast as the models change. Think of it as something loosely modeled on FINRA, the industry-run regulator that oversees Wall Street brokerages.
The interesting part wasn't just the pitch. It was who agreed with it in the room.
Sam Altman, sitting at the same table, told the assembled world leaders: "Do not cede your responsibilities to AI labs like mine." Anthropic's Dario Amodei has previously said he'd "prefer Hassabis's slower timeline." Getting the CEOs of OpenAI, Anthropic and Google DeepMind on record wanting the same thing in the same week is rare enough to notice.
The problem is that the government Hassabis wants to lead this watchdog is moving in a very different direction.
Earlier this year, the Trump administration rebranded the US AI Safety Institute as CAISI, the Center for AI Standards and Innovation. The word "safety" was dropped on purpose. CAISI still runs voluntary pre-release evaluations of frontier models from OpenAI, Anthropic, Google DeepMind, Microsoft and xAI, and signed an MOU with GSA in March to help vet AI used across the federal government. But it has no enforcement power. Labs opt in. There is no published rule that requires anyone to slow down, change course, or stop shipping a model if CAISI finds something alarming.
That gap is what Hassabis is essentially asking to close. And it's a bigger ask than it sounds:
- FINRA is not small. The financial regulator he's citing as a model oversees roughly 4,100 firms and 636,000 registered brokers, with a budget close to $1 billion and around 3,400 employees. A FINRA-for-AI would be a real institution, not an advisory board.
- The resource gap is already there. Paul Lekas of the SIIA pointed out to the Senate that the UK's AI Security Institute has roughly 10 times the staff and funding of its US counterpart. If CAISI is going to grow into what Hassabis is describing, someone has to pay for it.
- The kill switch already sort of exists. Less than a week before the G7 lunch, the Trump administration used export controls to force Anthropic to cut off access to its most advanced model in certain markets. Nobody at the lunch brought this up. Washington already has levers. Hassabis is asking to formalize them.
There's also a version of this that isn't about safety at all. A regulator that requires expensive pre-release testing and world-class technical reviewers would, by design, be easier for well-funded incumbents to comply with than for smaller labs or open-source projects. Hassabis, Altman and Amodei run three of the four best-resourced AI companies on earth. Asking for a regulator you can afford is not exactly a sacrifice.
Still, none of this is happening in a vacuum. Protesters marched on OpenAI, Anthropic and Google DeepMind's San Francisco offices last week demanding the labs slow down. The UN Secretary General called for a global ban on AI-powered weapons in the same news cycle. The pressure for someone to be in charge of this technology is coming from everywhere at once. The labs would rather that someone be an industry-friendly technical body they helped design.
The gap between what the frontier CEOs are asking for and what the White House is doing is the actual story here. When the people building the most powerful AI models are publicly asking a government to regulate them harder, and that government responds by removing the word "safety" from the name of its main oversight body, someone is going to end up disappointed. If Hassabis is right that a real watchdog is coming, the question is just whether it gets built in Washington, Brussels or Beijing. Right now, Washington doesn't seem interested in the job.
