OpenAI, Anthropic, and Google Form AI Safety Working Group for Shared Testing Standards

The three leading AI laboratories have been quietly collaborating since July 2026 to establish coordinated safety protocols and testing frameworks, signaling an unprecedented level of cooperation among competitors. Representatives from OpenAI, Anthropic, and Google DeepMind have been meeting as a working group to develop shared technical evaluation standards and pre-release auditing procedures for frontier AI models, according to reporting in mid-September 2026.