Anthropic, OpenAI, and Google DeepMind have been holding quiet working-group meetings since July to explore creating an industry-led AI standards body, The Information reported on September 14, 2026, with The Washington Post and multiple outlets confirming the story hours later.

This is the first time the three dominant frontier AI labs have coordinated on governance at this level. For businesses that run or are planning to run AI systems, this is a signal worth paying attention to — the compliance landscape for AI is about to get more structured, whether through industry agreement or government action.

What They’re Actually Proposing

The three CEOs are not in full agreement on what the body should look like, but the core purpose is consistent: establish shared protocols for testing frontier AI models before they are released to the public.

That means independent evaluations, pre-release safety reviews, and standardized risk assessments — the kind of audit process that currently does not exist in any formal way.

Dario Amodei (Anthropic) is driving the push. His model is an “FAA for AI” — a federal agency with the authority to block a model’s release before it goes public. Amodei has written that powerful AI could deliver enormous economic benefits while also creating risks that demand stronger safeguards, and his essay “We Must Pace the Frontier” laid out the framework he is now pushing inside these discussions.

Demis Hassabis (Google DeepMind) is proposing something closer to a “FINRA for AI” — an industry-funded, federally overseen standards body that starts with voluntary pre-release review. This is a softer approach than Amodei’s, but still represents a major shift from the current situation where labs largely self-regulate.

Sam Altman (OpenAI) backs an independent labs-led body in the absence of federal action. Altman told Fortune this week that OpenAI will not go public in 2026, citing safety concerns as a reason to stay private for now.

All three want to move before the end of 2026. Whether they can agree on the structure is the open question.

Why This Is Happening Now

Two dynamics converged to push this forward.

First, the government has not moved. The AI Kill Switch Act and other proposed federal legislation have stalled. The midterm regulatory environment in 2026 has produced more positioning than actual law. The labs are filling the vacuum rather than waiting for Congress to act.

Second, capability has outpaced trust. Anthropic’s Fable 5.1 model currently leads benchmarks for autonomous tasks. AI agents are now completing multi-week work cycles without human checkpoints. The reliability gap that enterprise teams have been talking about for two years is closing fast — which means the governance gap is becoming more urgent, not less.

When AI can act autonomously over weeks and handle consequential business decisions, the question of who is responsible for testing it before deployment is not abstract. It is a procurement question, a legal question, and a board-level risk question.

The Three Models on the Table

Model	Analogy	Who Prefers It	How It Works
Federal agency with blocking power	FAA	Amodei (Anthropic)	Agency can halt a model release before it ships
Industry-funded, federally overseen	FINRA	Hassabis (Google)	Labs fund it, government sets minimum standards
Independent labs coalition	No direct analogy	Altman (OpenAI)	Labs agree to standards without government mandate

The difference between these models matters significantly for enterprise buyers. A FINRA-style body would produce a certification standard that procurement teams could reference. An FAA-style body would mean no frontier model ships without independent sign-off. A voluntary coalition produces guidance but no enforcement.

What This Means for Business

For most businesses running or evaluating AI systems, this week’s news has three practical implications.

Vendor selection is a governance decision. The labs building the tools you deploy are now explicitly on record about their approach to safety and testing. How seriously a vendor takes pre-release evaluation is now a reasonable question to ask in a procurement process — not just a technical question but a risk management question.

Compliance requirements are coming. Whether the standards body ends up being mandatory or voluntary, the direction is toward more formal audit requirements for frontier AI systems. Businesses that build AI governance into their operations now will have a lighter lift than those that treat it as an afterthought.

The pace of deployment still matters. None of this represents a slowdown in what AI can do for your business. The discussion is about standards for the labs, not restrictions on enterprise use. If anything, a more structured safety regime should increase confidence that the tools being deployed have been evaluated before reaching your operations.

For businesses thinking about where to place their AI bets, the message is straightforward: the companies building AI are taking accountability seriously enough to coordinate with competitors on it. That is a different level of institutional commitment than the industry showed two years ago.

Enterprise DNA’s view has always been that sustainable AI adoption is built on foundations that can be governed — agents with defined roles, oversight built in from the start, and measurement that makes ROI visible. The industry is slowly arriving at the same conclusion.

---

Building an AI agent workforce for your business? Enterprise DNA’s Omni team works with business owners to deploy AI agents that fit real operations, with the governance and oversight built in from day one.

Book a discovery call with the Omni team