# OpenAI, Anthropic And Google Want To Write Their Own AI Rules. *AI · 15 September 2026 · RECORDx NEWS* Watch: https://www.youtube.com/watch?v=s1WJiRBeGoM Page: https://news.recordx.co/story/openai-anthropic-and-google-want-to-write-their-own-ai-rules/ On Sunday 13 September 2026, The Information reported that Anthropic, OpenAI and Google have been holding working-group meetings since July about setting up an industry-led standards body for AI. What it would do is test and audit frontier models. In some versions of the plan, before those models are released. They agree they need a referee. They do not agree on what it should be. Anthropic CEO Dario Amodei is driving the effort, focused on technical testing and pre-release auditing. His model has been described as an FAA for AI: a body with the power to block a model's release. On 12 September he published a post proposing that third-party watchdogs be embedded inside the labs. Google DeepMind's Demis Hassabis published a proposal in July for a FINRA for AI: industry-funded, federally overseen, starting with voluntary pre-release reviews that could harden into mandatory market-access rules. FINRA is the private, industry-funded body the securities industry built instead of being regulated directly. OpenAI's Sam Altman has said he favours a testing and auditing organisation, but told a company town hall he believes the labs will have to build it themselves, without the backing of the US government. OpenAI chief scientist Jakub Pachocki has said "shared safety standards and international coordination on further AI development need to be priorities now." Not everyone wants it built. Meta CEO Mark Zuckerberg advised President Trump against the idea. Meta is not in the working group. Nothing has been founded. There is no charter and no commitment — these are conversations, reported, not announced. Asked about them, all three companies declined to comment or did not respond. ## The spine The turn: they agree they need a referee and disagree on what it is. And the one thing none of them would do is talk about it — all three declined to comment or did not respond. ## Who wants what Demis Hassabis (Google DeepMind) published a proposal in July for a FINRA for AI — industry-funded, federally overseen, starting with voluntary pre-release reviews that could harden into mandatory market-access rules. ## On the record "I believe that shared safety standards and international coordination on further AI development need to be priorities now." — Jakub Pachocki, chief scientist, OpenAI Altman has said the companies "look forward to collaborating with our colleagues across the industry to formulate the best version" of such standards. ## The opposition Meta CEO Mark Zuckerberg advised President Trump against the idea. Meta is not in the working group. ## Context that makes it sharp FINRA is the model on the table: a private body, funded by the industry it polices, overseen by a federal regulator. It is what the securities industry built instead of being regulated directly. In July 2026 Treasury Secretary Scott Bessent floated a FINRA-like agency in a different context. A former Anthropic researcher resigned on 9 September. All three companies declined to comment or did not respond to requests. ## What it is not This is a set of conversations, not an agreed body. Nothing has been founded, no charter exists, and no company has committed. Reported, not announced. ## References Reporting drawn from: The Information, CNN Business, PYMNTS ## Transcript Three companies build most of the frontier AI in the world: Anthropic, OpenAI, and Google. Since July, they have been quietly meeting about the same question: Who should be allowed to referee them? The Information reported it on Sunday. Working group meetings running for two months on setting up an industry-led standards body for artificial intelligence. What it would do is test and audit frontier models. In some versions of the plan, before those models are released to the public at all They agree they need a referee. They do not agree on what it is, and the differences are not small. Dario Amodei of Anthropic wants something close to an FAA for AI, a body with the power to stop a model shipping. This weekend, he proposed embedding outside watchdogs inside the labs Demis Hassabis of Google DeepMind wants a FINRA for AI. Industry-funded, federally overseen. Voluntary reviews at first that could harden into rules you cannot sell without Sam Altman of OpenAI wants the testing organization too. He told a company town hall he thinks the labs will have to build it themselves without the United States government behind it Not everyone wants it built. Mark Zuckerberg of Meta advised President Trump against the idea. Meta is not in the working group, and when reporters asked, all three companies declined to comment Nothing has been founded. There is no charter and no commitment. What there is, is the three companies that build the thing drafting the rules for the thing ## Related - [His Plane Sank. He Swam Out And Waited In The Lake.](https://news.recordx.co/story/his-plane-sank-he-swam-out-and-waited-in-the-lake/) - [He Quit Anthropic To Warn About AI. Its Safety Lead Agreed.](https://news.recordx.co/story/he-quit-anthropic-to-warn-about-ai-its-safety-lead-agreed/) - [They Tried To Take Off Again. The Wheels Were Already Down.](https://news.recordx.co/story/they-tried-to-take-off-again-the-wheels-were-already-down/) - [New York Said The Air Was Safe. Its Own Files Say Otherwise.](https://news.recordx.co/story/new-york-said-the-air-was-safe-its-own-files-say-otherwise/)