Google, OpenAI and Anthropic are discussing a self-regulatory AI safety body, reportedly targeted for late 2026 or early 2027. Proposed rules include pre-release third-party tests, incident reporting and checks on voluntary pledges. The body has not launched.
The effort originally sought a public-private partnership with federal oversight, but stalled after a draft White House executive order was put on hold and the administration told the companies to reach industry consensus first. Modeled on FINRA, SAFA could set standards but may lack power to enforce them without government registration.
No charter or leader has been confirmed. The CEO shortlist includes Sriram Krishnan and Arati Prabhakar; Condoleezza Rice and David Friedberg were separately shortlisted for chair. Meta, xAI and Nvidia pushed back publicly at Dreamforce on Sept. 15. Critics say smaller competitors and open-source developers would face rules they did not help set.
