OpenAI and Anthropic Conducted Mutual AI Model Safety Testing in August 2025
OpenAI tested Anthropic's Claude models while Anthropic evaluated OpenAI's GPT variants in a joint safety evaluation.
Mutual Model Testing Agreement
OpenAI and Anthropic negotiated a legal agreement to test each other’s AI models, according to reporting from The Information. The two leading AI labs conducted a joint safety evaluation in August 2025, marking a notable instance of direct collaboration between competitors in the AI safety space.
What Each Company Tested
During the evaluation, OpenAI tested Anthropic’s Claude Opus 4 and Claude Sonnet 4 models. Meanwhile, Anthropic tested OpenAI’s GPT-4.0, GPT-4.1, and other models during the same period.
It remains unclear if OpenAI and Anthropic finalized their mutual model-testing agreement following the August 2025 evaluation.
Anthropic’s Separate Testing Partnership
In mid-September 2026, Anthropic announced a separate partnership with Accenture focused on independent evaluation of frontier AI models. The two organisations expect to invest over $1 billion in building AI model testing capacity over the next five years.
Industry Perspectives on AI Safety Coordination
Nvidia CEO Jensen Huang stated that new antitrust laws or regulations to allow AI companies to coordinate safety work are “completely unnecessary.” Meta CEO Mark Zuckerberg asserted that every lab has the responsibility and incentive to move at its own pace to train models safely.