The Regulator They Built for Themselves: Google, OpenAI and Anthropic Court Sriram Krishnan to Lead a 'Frontier AI Standards Agency'

The three companies that define the frontier of artificial intelligence have decided they know best who should police it: themselves. According to a report by The Information on September 24, 2026, 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 proposed self-regulatory body tentatively called the Frontier AI Standards Agency.

The irony at the center of the story is thick enough to cut with a knife. Krishnan served in the second Trump administration from January 2025 until June 2026, and he spent much of that tenure arguing publicly against exactly the kind of institution he is now being asked to run. On his way out of the White House he declared that “there will not be an FDA for AI,” insisting that a centralized agency requiring “a team of lawyers before you can get a model out” would put “sand in the gears” of the AI revolution. Now the industry wants him to build the gearbox.

How the deal took shape

The initiative has been taking shape in stages for months. Google DeepMind chair Demis Hassabis initiated the talks in July 2026, publishing a governance proposal for a U.S.-led oversight body that could vet cutting-edge models and — controversially — trigger industrywide slowdowns when credible threats emerged. At the time, Krishnan, still in office, dismissed the idea, telling reporters that an “FDA for AI” would not be created within the administration.

The proposal accelerated in September. 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. Lehane argued that no antitrust waiver was needed for the arrangement and endorsed a provision of the pending FRONTIER Act that would require “independent verification organizations” inside frontier labs. OpenAI has since said publicly that it will proceed with the body “with or without government support.”

The body’s design borrows deliberately from Wall Street: it is modeled on FINRA, the financial industry’s self-regulatory organization, which writes and enforces rules for broker-dealers under federal oversight. The target launch window is late 2026 or 2027 — with no government oversight attached.

What the agency would actually do

The scope under discussion is more concrete than most trial balloons:

• Assessment protocols — shared technical evaluations for frontier models
• Pre-release audits — third-party safety testing before model deployment
• Incident-reporting rules — standardized disclosure when models misbehave in the wild
• National-security testing — evaluation partnerships with federal agencies and the U.S. National Labs

Participating vendors would be encouraged to publish detailed model cards, maintain strong internal cybersecurity, vet key personnel, and adequately fund safety and security research. On paper, it reads like a checklist of everything critics have spent three years asking frontier labs to do voluntarily.

“A cartel by any other name”

Not everyone is convinced. Cohere CEO Aidan Gomez has called the proposal “a cartel by any other name,” drawing parallels to the SEC’s 1975 designation of Nationally Recognized Statistical Rating Organizations — the regime that anointed a handful of credit-rating agencies whose failures later lubricated the 2008 financial crisis — and to Europe’s 1985 Motor Vehicle Block Exemption. His warning is straightforward: a standards body controlled by the founding three would entrench incumbents, raise barriers for smaller competitors, and squeeze open-source developers who cannot afford participation.

The governance questions are equally sharp. Who funds the agency, and does “verifier” status require paying the founding three? What formal role, if any, do NIST or the National Labs get in the testing pipeline — or does the industry simply rent their credibility? Anthropic and Google, notably, have not publicly confirmed their participation to reporters, leaving the initiative in what observers have described as strategic ambiguity: if the body flops, it can be remembered as an OpenAI-led trial balloon; if it succeeds, everyone was a founder.

A crowded week for AI governance

The announcement lands in the same news cycle as several competing answers to the question of who watches the frontier. Mark Carney and Emmanuel Macron have been pushing for a global Technology Stability Board with central-bank-style legitimacy. Twenty-six state attorneys general told Congress this week to regulate AI now or watch the states do it themselves. Senator Bernie Sanders and Representative Greg Casar introduced a bill that would outright ban superintelligence. And the White House — having just asked OpenAI and Anthropic to hold models back from the UK AI Security Institute until American reviewers finish their own checks — is simultaneously signaling that it wants U.S. review to happen on U.S. terms, by whatever institution gets there first.

Against that backdrop, a FINRA-for-AI staffed by the labs themselves is the industry’s bid to define the terms before statute does. The editors at AI Weekly frame the stakes plainly: a self-regulator run by the person who argued Washington should not run one is either the cleanest way to head off statutory rules or a legitimacy problem in waiting — and any lab outside the founding three should press for governance seats before the terms of reference are locked.

The two signals to watch before the end of the year are Krishnan’s answer, and whether Anthropic or Google confirm participation on their own terms rather than letting OpenAI carry the flag alone. A founding charter with three names on it would mark the most significant structural commitment to self-regulation the industry has ever made. An OpenAI-only announcement would mark something else: the moment the frontier’s regulators stopped pretending to be neutral.

For now, the most powerful sentence in the story belongs to Krishnan’s own past: “There will not be an FDA for AI.” The industry is betting roughly a trillion dollars of market capitalization that he was right — and that a FINRA will do instead.