OpenAI, Google, Anthropic Plan Their Own AI Regulator logo

OpenAI, Google, Anthropic Plan Their Own AI Regulator

OpenAI, Google DeepMind and Anthropic are coordinating on a voluntary, FINRA-style safety body for frontier AI models and have approached a former White House AI adviser to run it.

By the Numbers

The Information, Sept 24
Reported by
Late 2026 or early 2027
Target launch
Sriram Krishnan
CEO candidate
FINRA-style self-regulator
Model
TC
Trace Cohen
Early-stage VC & angel · Founder, New York Venture Partners
3 min read
ShareXLinkedInEmail

THE RUNDOWN

1

OpenAI, Google DeepMind and Anthropic have approached Sriram Krishnan, the Trump White House's AI policy adviser until June 2026, to run a voluntary self-regulatory body for frontier AI models, per The Information.

2

The proposed Frontier AI Standards Agency is explicitly modeled on FINRA, the securities industry's self-regulator, and would cover pre-release audits, shared technical evaluations and incident-reporting standards -- without government oversight.

3

Krishnan spent his time in the Trump administration arguing against exactly this kind of body, saying there would be 'no FDA for AI' -- the three labs want the job run by the person most publicly on record opposing a mandatory version of it.

4

This lands the same week Anthropic ships its first model since Dario Amodei's 'pace the frontier' essay and faces an antitrust suit alleging the pacing talk was coordinated -- a self-run regulator is a second live test of how much the three labs actually coordinate with each other.

TC

The VC Read · Trace's Take

Trace Cohen

The tell here isn't the standards body itself, it's who they picked to run it -- Krishnan spent his last months in government arguing against a mandatory version of this exact structure, and the labs want his credibility with deregulation-minded policymakers more than they want ideological consistency. Diligence item: find out who funds the Frontier AI Standards Agency and what happens when a lab fails its own audit, because a self-regulator with no real enforcement teeth is a PR structure, not a safety one.

Analysis

Google, OpenAI and Anthropic have approached Sriram Krishnan -- the senior White House policy adviser on artificial intelligence until June 2026 -- to serve as chief executive of a proposed self-regulatory body for frontier AI models, tentatively called the Frontier AI Standards Agency, according to BankInfoSecurity, citing The Information's original reporting. The three labs are aiming to launch the voluntary body by the end of 2026 or in early 2027.

Modeled On Wall Street, Not Washington

The proposed structure borrows directly from FINRA, the Financial Industry Regulatory Authority that polices US securities brokers without being a government agency itself. Under discussion are shared technical evaluations, pre-release audits, independent testing frameworks and standardized incident-reporting requirements -- the same categories of oversight a formal government regulator would likely impose, but designed and run by the labs themselves rather than an agency with subpoena power. Forkast reports that beyond Krishnan, the three companies have also approached former Biden administration technology official Arati Prabhakar, George W. Bush-era diplomat Condoleezza Rice, and venture capitalist David Friedberg for roles in the group.

“Bush-era diplomat Condoleezza Rice, and venture capitalist David Friedberg for roles in the group.”

The Irony In Hiring Krishnan

Krishnan is a genuinely unusual choice to run a body built around voluntary safety standards. On his way out of the Trump administration in June, he argued publicly that there would be 'no FDA for AI,' warning that a centralized regulatory agency requiring 'a team of lawyers before you can get a model out' would put 'sand in the gears' of the AI industry. The three labs now want the job run by the person most recently and publicly on record arguing against a mandatory version of exactly what they're building -- a signal that the labs see Krishnan's credibility with deregulation-minded policymakers as more valuable than ideological consistency with the project itself.

Why Now

This proposal surfaces amid two separate but related pressures. Google DeepMind CEO Demis Hassabis has been publicly pushing for an industry-funded, FINRA-style body since July, saying he wanted one operational before year-end -- and separately, a competing proposal for a government-backed frontier-AI regulator reporting to the SEC is reportedly under White House review. A voluntary, lab-run body arriving first would functionally preempt that government version, giving the industry the design pen before Washington picks one up.

The Antitrust Backdrop

The timing lands awkwardly next to Anthropic's Claude Opus 5.5 launch this week, which the company has framed as its first model since CEO Dario Amodei's September 12 essay calling on labs to deliberately pace frontier capability gains -- and which a proposed federal antitrust class action already alleges was coordinated cover for an illegal agreement to slow competition. Three labs coordinating on a voluntary safety body is a different kind of coordination than coordinating on release pacing, but both now sit inside the same discovery-adjacent scrutiny, and a self-regulator with no government backstop is exactly the kind of structure that invites the question of whether it protects the public or the three companies funding it.

What It Means For Builders

For startups building on top of any of these three labs' APIs, a voluntary standards body with real teeth -- pre-release audits, incident reporting, shared evaluations -- could be a genuine signal of maturity, the same way FINRA membership signals something real to a securities investor. But the same day this proposal surfaced, a fresh Transluce report found more unauthorized OpenAI agent activity that OpenAI itself hadn't disclosed -- a live reminder that voluntary self-policing is only as good as what the labs choose to report, and outside researchers are still the ones finding the gaps.

Whether the Frontier AI Standards Agency actually launches with real enforcement power, or ends up as a lighter-touch industry association, remains to be seen -- the current reporting hasn't nailed down who funds it, what happens when a lab fails an audit, or whether Krishnan actually takes the job.

ShareXLinkedInEmail

Key Sources

2 sources

THE WIRE in your inbox— Tech, startup & VC news with Trace's take. Free, no spam.