Frontier artificial intelligence developers Google, OpenAI, and Anthropic are advancing plans to establish a self-regulatory oversight authority to test and audit advanced models without government supervision.
Provisionally named the Standards Authority for Frontier AI (SAFA), the initiative aims to launch by late 2026 or early 2027 to build standardized safety benchmarks for cutting-edge models. Under the proposed blueprint, the entity would support pre-deployment technical testing, establish unified safety incident reporting protocols, and qualify independent third-party auditors. Participants are also debating whether the body should directly conduct capability evaluations alongside federal efforts such as the Center for AI Standards and Innovation.
The momentum behind the initiative originated from an idea put forward in July by Google DeepMind co-founder Demis Hassabis, who suggested a self-regulatory framework inspired by Wall Street's Financial Industry Regulatory Authority. Although the companies initially hoped for a public-private partnership backed by federal oversight, they shifted toward an independent industry structure after related executive action in Washington stalled. The consortium has reportedly approached former White House artificial intelligence policy adviser Sriram Krishnan and other notable figures to discuss executive leadership roles.
The project has nonetheless drawn skepticism from competing AI developers, who caution that voluntary coordination among tech titans could consolidate power into an exclusionary cartel. As public debate intensifies, the participating labs maintain that private technical coordination is necessary to safeguard against catastrophic hazards while government frameworks struggle to keep pace.