Stories
OpenAI, Anthropic Eye Landmark AI Safety Pact for Mutual Model Testing
M
Moneycontrol•21-09-2026, 22:35
OpenAI, Anthropic Eye Landmark AI Safety Pact for Mutual Model Testing
- •OpenAI and Anthropic are reportedly negotiating a legally binding agreement to stress-test each other's commercial AI models.
- •The proposed pact involves mutual API access to probe for vulnerabilities, with a guarantee against data retention.
- •This initiative follows disclosures of serious internal safety breaches at OpenAI and growing calls for stronger AI oversight.
- •Anthropic CEO Dario Amodei advocates for pacing AI development with stronger safety testing and international cooperation, a sentiment echoed by OpenAI CEO Sam Altman.
- •Leading figures like Elon Musk, Mark Zuckerberg, Jensen Huang, Mustafa Suleyman, Bill Gates, and various governments have shared diverse views on AI safety, regulation, and potential risks.
Key Questions Answered
OpenAI and Anthropic are negotiating a legally binding agreement to provide each other with API access to commercial artificial intelligence models to probe for vulnerabilities, with a strict guarantee that neither side will retain the other's data.
Anthropic CEO Dario Amodei proposed a three-step plan to pace AI development, which includes stronger safety testing, independent evaluations of advanced models, international cooperation, and coordination among leading AI companies.
OpenAI intends to adopt Anthropic's proposal by committing to independent evaluators with employee-like access and will begin regularly publishing reports on unexpected or unauthorized AI behavior.
