What happened
The three leading AI labs are negotiating the formation of a self‑regulatory body, tentatively called the Frontier AI Standards Agency (also referred to as the Standards Authority for Frontier AI or SAFA). The plan, first reported by The Information on September 24, proposes that the agency set safety standards and incident‑reporting protocols for the most advanced AI models. OpenAI’s head of global affairs, Chris Lehane, publicly confirmed on September 15 that the companies are coordinating on the effort. The working name and the potential appointment of Sriram Krishnan – a former senior White House adviser on AI who publicly opposed an FDA‑style licensing regime – come from anonymous sources cited by The Information. Google and Anthropic have not yet issued independent confirmations. The agencies aim to emulate FINRA’s model, creating industry‑wide rules and enforcement mechanisms without direct government authority.
According to The Information, talks began in July after DeepMind chair Demis Hassabis suggested a US‑led body to evaluate frontier AI models and trigger industry‑wide slowdowns when threats emerge. The proposed agency, referred to as Frontier AI Standards Agency or SAFA, would be industry‑run rather than a government entity.
OpenAI’s Chris Lehane confirmed on September 15 that the three companies are coordinating on the effort, marking the only publicly verified element of the plan. The rest of the details – including the agency’s name, structure, and the potential appointment of Sriram Krishnan as chief executive – rely on anonymous sources.
Krishnan, who served as a senior White House adviser on AI from January 2025 to June 2026, left the administration after publicly rejecting a licensing regime for AI models. His potential leadership is seen as a way to give the agency credibility in Washington while maintaining industry control.
Source details: pasqualepillitteri.it ↗
Why it matters
A joint industry standards body could shape how the most powerful AI systems are developed, tested, and deployed, potentially filling a regulatory gap that governments have struggled to address. By establishing incident‑reporting deadlines and shared safety protocols, the agency could reduce the risk of high‑impact breaches like the recent OpenAI Medicare portal incident and Anthropic’s Claude attacks. The involvement of a former White House adviser lends political credibility, but also raises concerns about self‑regulation and possible barriers for smaller firms and open‑source developers. If the agency’s charter grants it enforcement powers similar to FINRA, its standards may become de‑facto requirements for participating companies, influencing market dynamics and future policy discussions.
The agency would address a critical gap in by establishing standardized incident‑reporting timelines and safety benchmarks, which could mitigate the fallout from future breaches similar to the OpenAI Medicare portal incident and Anthropic’s Claude attacks.
By modeling itself after FINRA, the agency could enforce compliance among its members, creating a unified set of rules that could become a for the broader AI market, influencing both domestic and international regulatory discussions.
Critics, such as Cohere CEO Aidan Gomez, warn that a standards body dominated by the three largest labs could act as a barrier to entry for smaller companies and open‑source projects, potentially consolidating market power and limiting competition.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Impossibility results in algorithmic fairness (e.g. Kleinberg et al., Chouldechova) show what?
What to watch next
Key milestones include the release of the agency’s charter – detailing governance, funding, voting rights, and enforcement powers – and formal confirmations from Google and Anthropic. Observers will also watch whether the agency adopts mandatory testing regimes, how it interacts with existing government initiatives, and whether it attracts participation beyond the three founding labs. The reaction of regulators, competitors, and open‑source communities will indicate the body’s broader acceptance and impact.
The publication of the agency’s charter, which will clarify who funds the body, how voting rights are allocated, and what enforcement mechanisms are available.
Official statements from Google and Anthropic confirming their participation and outlining any conditions or expectations they have for the agency.
Reactions from U.S. regulators, European authorities, and industry groups, especially regarding whether the agency’s standards will be recognized or integrated into existing legal frameworks.
Potential expansion of membership beyond the three founding labs, which could affect the agency’s perceived legitimacy and its ability to set broadly applicable standards.