Google DeepMind CEO Demis Hassabis is calling for a new standards body to review frontier AI systems before they are released, adding a concrete proposal to the debate over how advanced models should be governed.

In a Tuesday post titled "A Framework for Frontier AI and the Dawning of a New Age," Hassabis argued for a self-regulatory organization modeled in part on FINRA, the U.S. financial industry regulator. The body would be technically focused, staffed by industry experts and open source representatives, and funded by frontier AI labs.

The proposal would begin voluntarily. Frontier labs would share models with the standards body for review up to 30 days before release. If the assessment process proved effective, Hassabis said the model could be formalized so frontier models would need to pass review before deployment in the U.S. market. Labs would also work with the body on critical vulnerabilities discovered after release.

The idea sits between two current regulatory poles: government-led licensing regimes and purely internal safety evaluations by model developers. Hassabis framed the standards body as a way to keep pace with fast-moving technical risks while preserving room for innovation.

The proposal is not yet a binding policy or a government plan. Its significance is that one of the leading frontier AI labs is endorsing a more structured pre-release review process, at a moment when governments are still deciding how much direct oversight advanced model launches should face.