Google, OpenAI, Anthropic Reportedly Plan AI Safety Standards Body

Google, OpenAI, and Anthropic are reportedly in the midst of establishing an AI safety standards body. This coalition, tentatively called the Standards Authority for Frontier AI (SAFA), is expected to launch by late 2026 or early 2027. The proposed body aims to create independent safety standards without federal oversight, following a stalled proposal for federal regulation.

The three tech giants are finalizing plans for the body following discussions that began with a proposal by Google DeepMind co-founder and Alphabet chief scientist Demis Hassabis. The proposed group would focus on pre-deployment testing, incident reporting, and auditor qualifications. Its standards and enforcement powers have not been announced, but the group is considering in-house evaluations.

Former White House AI policy advisers and former officials are being considered for the chief executive and board positions. While self-regulation can establish a baseline for model behavior, concerns remain about the potential barriers to entry for smaller developers and the lack of a clear legal framework. For end users and enterprise customers, the creation of SAFA could impact the speed and reliability of new generative AI tools.

However, the body's credibility will depend on the independence of evaluators and the widespread adoption of its standards.