Google, OpenAI and Anthropic plan AI safety body

Google, OpenAI and Anthropic are planning an industry body to develop or support common safety assessments for advanced artificial intelligence models, according to reporting attributed to The Information. The proposed organization, tentatively named the Standards Authority for Frontier AI, could launch in late 2026 or 2027, but the companies have not publicly confirmed its structure, membership or authority. The reported plan is for the body to operate independently of government oversight and help establish shared approaches to evaluating frontier AI models. Reports describe possible work on risk assessment, testing and review before models are released. It remains unclear whether the organization would conduct assessments itself or support developers and outside evaluators. The launch window is an expectation, not a confirmed date. One account described a possible launch in late 2026 or 2027, while another specified early 2027. The plan’s details, including its governance and relationship to existing industry safety efforts, have not been announced.