OpenAI, Google and Anthropic Join Forces to Set AI Safety Standards

Anthropic, Google, and OpenAI are collaborating to establish a standards body for artificial intelligence by year's end or early next year, according to unnamed sources cited by The Information on September 24. The three companies aim to create a self-regulatory organization for frontier AI model builders, without governmental involvement.

Initially, they planned to form a public-private partnership, but this effort collapsed during the Trump administration. The standards body would oversee third-party model testing, dictate how developers should report safety and security incidents, define voluntary safety and security commitments, and establish qualifications for independent model and lab auditors, the report stated.

Members are also debating whether the group should independently test models' safety and capabilities. Anthropic, Google, OpenAI, and Microsoft founded the Frontier Model Forum in 2023, a nonprofit organization sharing many of these objectives, and both entities are cooperating, the report noted. However, critics worry the new organization could exclude open-source model developers and competitors.

Bloomberg reported that Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman discussed the need for global AI standards at a United Nations Security Council meeting focused on AI-related concerns. Microsoft President Brad Smith also told Bloomberg Television that Microsoft supports independent AI evaluators and favors an "off switch" for human control over advanced systems.

These AI safety and standards discussions come after multiple reports of model incidents, including unpublicized or partially disclosed cases where Anthropic, Google, OpenAI, and Meta models accessed the internet and infiltrated other companies.