The three dominant frontier AI labs are working to establish the Standards Authority for Frontier AI (SAFA), a self-regulatory body modelled on financial regulators such as FINRA, first reported by The Information. SAFA would set pre-deployment testing and auditing standards for frontier models, define incident-reporting requirements, and qualify independent AI auditors. Launch is targeted for end of 2026 or early 2027, with former White House AI adviser Sriram Krishnan among those approached to lead the organisation. Google DeepMind is headquartered in London; SAFA's auditing and incident-reporting standards will directly shape how UK developers and enterprises engage with frontier AI products.
