OpenAI, Google and Anthropic advance plan for industry-run AI safety standards body

PYMNTS reports that Google, OpenAI and Anthropic are working toward a safety standards organization for frontier AI, citing The Information on September 23. The body could launch in late 2026 or 2027 and would operate without government oversight. Its tentative name is the Standards Authority for Frontier AI, and the companies have not announced its members, powers or standards. The article notes that the three firms already belong to the Frontier Model Forum, which dates from 2023, and that the new body's separate role has not been explained.

The report places this beside a parallel diplomatic effort. On September 22, 22 international leaders backed a call started by Norway's Prime Minister Jonas Gahr Støre and Finland's President Alexander Stubb. It seeks pre-deployment testing, independent evaluations with model access, common standards and reporting of serious incidents, and asks UN members to look at creating a body able to set standards, check compliance and bring governments together once systems cross capability thresholds. It proposes such a regime but does not create one. The Wall Street Journal reportedly said Canada's Mark Carney answered Støre's invitation within 10 minutes.

OpenAI says it will pursue voluntary standards with other labs with or without government support, and CEO Sam Altman told the UN Security Council that national and international standards should be compatible and should not favor established labs. PYMNTS observes that an industry-founded body could give incumbents sway over how rivals are judged. President Trump told the UN General Assembly the United States rejects any globalist scheme of control over AI, and the US framework relies on voluntary testing. The article describes several competing routes to oversight that may converge on technical evaluations but not necessarily on enforceable decisions.