Google, OpenAI and Anthropic are working on a new industry-led organization intended to set common safety standards for frontier AI models, according to a report published Thursday.
The Information reported, in a summary carried by Techmeme, that the companies are considering launching the group in late 2026 or early 2027. The working name is the Standards Authority for Frontier AI, or SAFA.
The proposed body would support third-party testing of advanced models before deployment, develop standards for reporting safety and security incidents and spell out qualifications for independent model and laboratory auditors, according to the report.
The discussions are notable because the organization would be created by the companies themselves rather than by a government regulator. That could let the labs move faster on shared technical practices, but it also puts the credibility of the effort on questions such as independence, transparency and whether participation remains voluntary.
Google, OpenAI and Anthropic already participate in the Frontier Model Forum, an industry group created to share information about risks and best practices around advanced AI systems. The proposed standards body would go further by focusing on common testing and auditing practices.
The news comes as AI executives are publicly calling for stronger safety coordination while governments remain divided over how far regulation should go. At the United Nations this week, executives from OpenAI and Anthropic urged international cooperation on safeguards for increasingly capable systems.
A private standards body would not settle the broader argument over who should police frontier AI. But it would mark an attempt by the largest developers to turn broad safety commitments into a more formal set of shared practices.
