Google, OpenAI, and Anthropic are planning to establish a self-regulatory body for AI safety, tentatively named the Standards Authority for Frontier AI (SAFA). The companies aim to launch the organization by the end of this year or early next year, operating independently of government oversight.
The proposed organization intends to support third-party testing of models before deployment and establish protocols for reporting safety and security incidents. Additionally, SAFA plans to convert existing voluntary safety commitments into formal industry standards and define the necessary qualifications for independent auditors.
Internal discussions are currently underway regarding whether SAFA should directly manage model safety and capability testing. Some participants have expressed concerns that the US Department of Commerce’s AI Safety Institute may lack the resources required to conduct comprehensive evaluations of frontier models.
While the companies continue to discuss the operational framework for SAFA, their executives are publicly advocating for the creation of unified safety standards and improved international coordination. The specific government agencies that might eventually oversee these standards remain undetermined.
