# OpenAI Calls for Government Role in Frontier AI Safety Standards

By Riza Dagoc

Canonical URL: https://www.tokenpost.com/news/regulation/24651
Published: 2026-09-27T23:09:57.000Z
Updated: 2026-09-27T23:09:57.000Z
Section: Regulation

> The proposed Standards Authority for Frontier AI would develop shared testing and audit standards, but its name, structure and government role remain unconfirmed.

Google, OpenAI and Anthropic are considering a proposed AI safety standards group that could shape how advanced models are tested and audited, but its formation remains unconfirmed.

The organization is tentatively called the Standards Authority for Frontier AI, or SAFA. The name is a working description, and the group has also been described more generally as a frontier-AI standards body.

SAFA could develop shared testing, auditing and safety standards for advanced models. Possible responsibilities include testing models, setting standards, qualifying auditors or taking on a broader oversight role, but no final structure has been established.

Government involvement also remains unconfirmed. No public launch announcement or formal founding document has established the group’s timing, responsibilities or relationship with government agencies.

The proposal comes amid calls for common rules as governments consider tighter oversight of frontier artificial intelligence. It would be separate from the Frontier Model Forum, which Google, Microsoft, OpenAI and Anthropic formed on July 26, 2023.

The Frontier Model Forum’s stated goals include advancing technical evaluations and benchmarks and developing a public library of safety solutions, best practices and standards. The distinction between that organization and a possible SAFA has not been publicly explained.

The proposal also reflects a debate over whether AI safety requirements should be established by governments, industry or both. OpenAI said July 15, 2026, that frontier-AI governance should include documented safety frameworks, serious-incident reporting and independent audits.

Chris Lehane, OpenAI’s chief global affairs officer, said critical decisions about frontier safety should be made by democratic governments rather than solely by frontier AI companies.

OpenAI also said a federal testing framework under discussion would establish testing standards, timelines and processes for highly capable AI models. The proposal for SAFA would begin with an industry-led standards process, while the role of government has not been settled.

Google DeepMind maintains a Frontier Safety Framework listed as version 3.1 and dated April 17, 2026. The framework is a separate company safety initiative and does not establish that SAFA has been formed.

OpenAI and Anthropic are reviewing tens of thousands of incidents involving frontier AI models, including attempts to bypass safeguards, escape sandboxes and evade monitoring, as described in earlier coverage of AI safety incidents.

The next step for the proposed organization has not been announced. Its final name, membership, oversight role and launch timing remain unsettled.

## Links in this article

- [earlier coverage of AI safety incidents](https://www.tokenpost.com/news/technology/24500)