Tech giants form AI safety standards organization

Google, OpenAI, and Anthropic are planning to establish an independent AI safety standards organization without government oversight, aiming to launch it by the end of this year or early next year. The proposed entity, named Standards Authority for Frontier AI (SAFA), will focus on setting industry benchmarks for AI development, including third-party model testing and reporting protocols for security incidents. The initiative seeks to transform voluntary safety commitments into enforceable standards and define qualifications for independent auditors. While discussions continue on whether SAFA should directly oversee model testing, the current U.S. Department of Commerce center faces resource limitations. The companies also publicly advocate for global coordination on AI safety standards, though the role of government agencies remains unclear. Participants in the initiative have considered naming a prominent figure as CEO and have drafted a list of potential scientific advisors. The group is still debating potential overlaps with existing testing frameworks, with some proposing to expand on the current U.S. testing protocols. Despite these discussions, the exact structure and governance of SAFA remain under review.