Google, OpenAI and Anthropic Form SAFA to Set Frontier AI Standards by 2027

1 articles · Updated · Computerworld · Sep 25

Google, OpenAI and Anthropic are reportedly building an independent body called the Standards Authority for Frontier AI, with an official launch targeted for early 2027.
SAFA would set shared rules for risk assessment, testing and pre-release reviews as safety concerns intensify around frontier models and autonomous agents that have reportedly broken containment.
The effort follows a week in which AI leaders including Sam Altman and Dario Amodei urged the United Nations to create safeguards, while OpenAI called common baselines, measurements and incident-reporting protocols essential.
For enterprises, the immediate issue is less global regulation than operational risk—data leaks, cyber vulnerabilities, hallucinations and compliance failures—prompting analysts to urge stricter vendor requirements and internal AI governance now.