Google, OpenAI and Anthropic reportedly plan their own AI safety standards body

Anthropic, Google and OpenAI are working on a new standards organisation for frontier AI and hope to launch it by the end of this year or early 2027, The Information reported on 24 September, citing unnamed sources. The report was summarised by PYMNTS and Newsquawk; the companies themselves have not announced the body.

What it would do

According to the report, the organisation would:

• support third-party groups that test models before they are released;
• set out how AI developers should report safety and security incidents;
• define the labs’ voluntary safety and security commitments;
• set qualifications for independent auditors of models and labs.

A working group is also considering whether the body should test models’ safety and capabilities itself.

Self-regulation, not a public partnership

The labs first sought a public-private partnership, but that effort stalled under the Trump administration, the report says; the current plan is for model builders to regulate themselves without government oversight. The Frontier Model Forum — a non-profit set up in 2023 by Anthropic, Google, OpenAI and Microsoft with many similar goals — remains active and is in contact with the new effort.

Criticism

Critics worry that the three biggest labs could use such an organisation to box out open-source developers and other competitors.

Context

The report came a day after AI executives briefed the UN Security Council, amid a broader argument over AI rules. Nvidia CEO Jensen Huang said in a podcast released on 23 September that AI firms should not get exemptions from antitrust or product-liability law, Reuters reported. Earlier this month Anthropic CEO Dario Amodei had called for an antitrust waiver so that labs could cooperate on safety.