The three most powerful AI labs in America want to build their own regulator, and they want it run by the man who spent 18 months in the White House insisting Washington should never build one. Google, OpenAI and Anthropic have approached Sriram Krishnan, the former Senior White House Policy Advisor on Artificial Intelligence, to serve as chief executive of a new industry self-regulatory body, according to a report from The Information on September 24.
The organization is tentatively called the Frontier AI Standards Agency, though some accounts of the same effort refer to it as the Standards Authority for Frontier AI, or SAFA. Either way, the template is Wall Street: the body is explicitly modeled on the Financial Industry Regulatory Authority, or FINRA, the self-regulatory organization that polices broker-dealers under the supervision of the Securities and Exchange Commission. People familiar with the plans say it could launch as soon as the end of this year or in early 2027.
From Hassabis's Post to a CEO Shortlist
The idea traces back to July 14, when Google DeepMind CEO Demis Hassabis published a framework on X calling for a FINRA-style standards body to test frontier models before release. Under his proposal, labs would initially share models voluntarily for review up to 30 days before launch, with mandatory assessment to follow once the protocol proved itself. "The strength of this approach is it would be technically focused, while at the same time supporting innovation and incentivising responsible behaviour," Hassabis wrote.
Since then, representatives of the three companies have met regularly as a working group. According to BigGo Finance, the labs first pursued a public-private partnership under federal oversight, but a draft White House executive order that would have created such a body failed to win enough support inside the Trump administration and was effectively shelved before summer's end. The companies then pivoted to pure self-regulation. OpenAI Chief Global Affairs Officer Chris Lehane confirmed at a Washington briefing on September 15 that the three labs had been coordinating on safety protocols for several weeks, and said no antitrust waiver was needed.
The scope under discussion is concrete rather than aspirational: assessment protocols, support for third-party pre-deployment safety testing, incident-reporting rules, qualification standards for independent auditors, and testing with federal agencies and the U.S. National Labs in national-security domains. The working group is still debating whether the agency should run its own safety and capability evaluations; some members reportedly argue that the Commerce Department's Center for AI Standards and Innovation lacks the resources to evaluate frontier models comprehensively.
Krishnan is not the only name in play. Former White House science adviser Arati Prabhakar is also reported to be in the CEO candidate pool, while former Secretary of State Condoleezza Rice and investor David Friedberg have been floated for chair. METR founder Beth Barnes and Paul Christiano, recently named to the OpenAI Foundation board, are reportedly under consideration for scientific advisory roles.
An Unlikely Choice, or a Perfect One
Krishnan, a former Andreessen Horowitz general partner, served in the administration from January 2025 until June 2026. Shortly before he left, he dismissed the prospect of a new executive-branch AI regulator in blunt terms: "there will not be an FDA for AI." He argued that a centralized agency requiring lawyers to sign off before a model ships would put sand in the gears of the AI boom.
That record is precisely what makes him attractive to the labs. A FINRA-style body is, by design, the alternative to an FDA-style licensing regime: industry-funded, technically staffed, and nominally independent. Hiring the official most associated with rejecting federal licensing signals to Washington that this is a substitute for statutory regulation, not a stepping stone toward it.
Why It Matters
The plan lands in a hostile legal and competitive environment. On September 19, a proposed class action filed in federal court in Northern California accused Anthropic, OpenAI, SpaceXAI and Google of illegally agreeing to slow AI development, arguing that collective restraint reduces the value consumers get from paid subscriptions. A formal standards body jointly founded by three of those defendants will inevitably be read by plaintiffs as more evidence of coordination, whatever Lehane says about antitrust waivers.
Competitors are already making that case. Cohere CEO Aidan Gomez called the proposal "a cartel by any other name," warning that such a body would entrench incumbents and raise barriers for smaller labs and open-source developers. Meta's Mark Zuckerberg, whose company is conspicuously absent from the talks, rejected the premise outright in a recent NBC News interview: "I don't think that we need some kind of industrywide coordination." He argued that liability and competition already give each lab reason to slow down internally when it finds problems.
That leaves the agency with a legitimacy problem from birth. FINRA works because the SEC stands behind it and membership is mandatory for broker-dealers. A frontier AI equivalent with no federal backstop and only three founding members would set standards chiefly for the companies that wrote them. The 2023 Frontier Model Forum, which counted Anthropic, Google, Microsoft and OpenAI as members and launched a safety fund of more than $10 million, showed how easily industry safety bodies can fade into principle statements. This effort is more ambitious precisely because it wants to test and audit, which is also why it invites more scrutiny.
What to Watch
The first signal is whether Krishnan accepts, and on what terms; a CEO announcement or founding charter before year-end would confirm this is a live institution rather than a trial balloon. Watch, too, whether Google and Anthropic publicly confirm participation, who funds the agency, whether labs outside the founding three, including Meta, SpaceXAI and Microsoft, are offered governance seats, and what role, if any, federal bodies such as CAISI and the National Labs are given. Finally, expect the plaintiffs in the September 19 antitrust suit to cite any formal charter as exhibit A.
“There will not be an FDA for AI.”— Sriram Krishnan, Former Senior White House Policy Advisor on AI