Google, OpenAI, and Anthropic are moving ahead alone with an AI safety standards body, tentatively called the Standards Authority for Frontier AI, or SAFA, aiming to launch by early 2027, according to Leo Schwartz and Stephanie Palazzolo at The Information. The labs first wanted federal oversight, but a draft executive order stalled after Meta CEO Mark Zuckerberg and Nvidia CEO Jensen Huang pushed back, and the White House told them to find consensus first. SAFA would back outside testing of models before release, set rules for reporting safety incidents, and certify independent auditors, and the labs approached Trump AI adviser Sriram Krishnan about running it. Because it would cover only frontier labs at first, critics worry its bar could shut out rivals like open-source developers.
