The labs are building their own watchdog. Congress wasn't invited.

Wednesday, the CEOs of OpenAI and Anthropic stood before the UN Security Council and said AI needs stronger oversight. Thursday, reporting revealed that Google, OpenAI, and Anthropic had quietly agreed to write the rulebook themselves. The outfit has a name: the Standards Authority for Frontier AI, or SAFA. According to reporting by The Information, it is modeled on FINRA — the self-regulatory body that polices Wall Street stockbrokers — except this one comes with no government supervisor. It could launch by the end of 2026 or early 2027. In theory, SAFA would back third-party testing of models before they ship, write the playbook for reporting safety incidents, and set qualifications for independent auditors. Still undecided: whether it actually tests the models itself, and whether member labs have to accept findings they don't like. The origin story is worth noting. The labs originally wanted a public-private partnership with federal oversight. That stalled when the White House's draft executive order failed to gain traction. So the three biggest labs are going it alone — Meta, DeepSeek, and open-source developers not included. Critics already have a name for it: regulatory capture. The auditors are the audited. The students are writing the exam, grading it, and deciding whether the answers count. The future isn't ungoverned. It's governed — by the three companies that had the most to lose from anyone else doing it. Sources: The Information (reported Sept 24, 2026, via TechRepublic Sept 25 and TechDefused Sept 26); AI Frontier Post; The Leadsman.