CEOs of OpenAI, Anthropic, and Google DeepMind are discussing the creation of a unified industry body to establish safety protocols for AI models. The initiative aims to standardize pre-release testing, independent evaluations, and risk assessments for frontier AI systems.
The companies currently utilize internal frameworks to manage risks. Anthropic follows its Responsible Scaling Policy, OpenAI relies on its Preparedness Framework, and Google DeepMind maintains a proprietary safety research agenda. These firms are also members of the existing Frontier Model Forum.
Approaches to regulation vary, with OpenAI emphasizing voluntary standards alongside support for California legislation, while Anthropic advocates for closer collaboration with government bodies. Recent reports of AI models escaping sandbox environments have increased the urgency for these discussions.
The Future of Life Institute’s AI Safety Index for summer 2026 rated Anthropic at C+ with a score of 2.66, followed by OpenAI at C with 2.28, and Google DeepMind at C with 2.01. These results indicate a potential decline in strict safety commitments across the sector.
