While the debate over who should set AI's rules rages, the three most powerful labs have quietly begun building their own referee. According to The Information, OpenAI, Anthropic and Google DeepMind have held regular working-group meetings since July to create an industry-led AI standards body — one that would independently audit the most powerful models before release, modeled on FINRA, the self-regulatory organization overseeing the US securities industry.
On September 15, Chris Lehane, OpenAI's chief global affairs officer, publicly confirmed the coordination for the first time at a Washington briefing. He said the three firms have been coordinating on safety for weeks and "do not need an antitrust waiver" — arguing that safety coordination is fundamentally different from commercial collusion. The same day, Google DeepMind launched the DeepMind Institute, whose first essays include Hassabis again calling for a US-led body that tests frontier models before deployment.
According to people familiar with the matter, the proposed framework includes shared technical evaluations, pre-release audits, independent testing and standardized safety protocols — and could require the most advanced models to undergo up to 30 days of review before public release. The three firms agree on the need for pre-release scrutiny but disagree on who should hold the pen: Anthropic favors a federal-agency model (like the FAA), DeepMind a federally overseen industry body (the FINRA model), and OpenAI an international certification forum (like the IAEA).
The opposition is equally loud. Cohere CEO Aidan Gomez publicly branded it: "This is a cartel." He argues that three of the largest labs writing standards and auditing each other amounts to building entry barriers in the name of "safety," suppressing competitors. Meta and xAI are also reported to be skeptical — neither participates in the working group. Critics further warn of "regulatory capture": when the regulated become the regulators, standards inevitably tilt toward the powerful.
Lehane's response: "It's better to try to work together to prioritize safety." But whether this "industry self-regulation" leads to safety or monopoly will only be clear once the standards body's charter is published. What's certain is that the next war of AI governance has begun around "who writes the rules" — and this time, the giants have chosen to move before the regulators.