The Information quotes sources as reporting that Google, OpenAI, and Anthropic are pushing forward to establish an industry self-regulatory organization named Standards Authority for Frontier AI (SAFA), which will operate independently without government regulation. The new organization is expected to be established by the end of this year or in early 2027 at the earliest.

The organization will focus on specific testing and auditing frameworks, including supporting security testing before third-party deployment, accident reporting regulations, and auditor qualification standards. This concept originated from a suggestion made in July by Google DeepMind, the Executive President, Demis Hassabis.

Initially, three companies sought to establish a public-private cooperation mechanism under federal regulation in the United States. However, after a draft executive order from the White House was put on hold, they turned to promoting pure industry self-regulation instead. The candidates for the executive president of this organization include former White House artificial intelligence policy advisor Sriram Krishnan. Nevertheless, whether SAFA will gain widespread recognition within the industry and whether the division of responsibilities with government regulatory agencies can be clarified remains to be seen.