Demis Hassabis Proposes U.S. AI Standards Body, Altman and Nadella Back It
The proposal represents a rare public alignment among leading AI CEOs on the need for external oversight of frontier models, with a concrete mechanism for testing and enforcement.
Reporting from 1 source: GIGAZINE.
Google DeepMind CEO Demis Hassabis has proposed a U.S. standards body to evaluate frontier AI models for dangerous capabilities. The plan, outlined in a July 14 document, would initially be voluntary and later mandatory. OpenAI CEO Sam Altman, Microsoft CEO Satya Nadella, and Google CEO Sundar Pichai expressed support. Hassabis warned that AGI could arrive within a few years and that competition has outpaced safety understanding.
In a document titled 'A Framework for Frontier AI and the Dawning of a New Age,' published July 14, 2026, Hassabis argued that artificial general intelligence could be achieved within a few years. He proposed a U.S. standards body modeled on the Financial Industry Regulatory Authority (FINRA) to test frontier models for risks including cyberattacks, biological threats, and loss of human control. The body would designate models as 'frontier-class' and require labs to publish model cards, maintain cybersecurity, and invest in safety research. Initially voluntary, the system would become mandatory after its effectiveness is confirmed. Hassabis said the framework would apply to any frontier-class model regardless of country of origin or whether it is open or closed.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.