OpenAI and Anthropic in talks to test each other's AI models for flaws

OpenAI and Anthropic are negotiating a legally binding agreement to test each other's commercial AI models for vulnerabilities, according to a report published Monday. The pact would grant each company API access to the other's models while guaranteeing data is not retained. The proposed mutual-testing agreement follows disclosures of internal safety breaches at OpenAI and growing calls for stronger oversight. Anthropic CEO Dario Amodei proposed a three-step plan for pacing AI development, which OpenAI CEO Sam Altman endorsed, while other tech leaders offered mixed reactions on regulation and safety. Industry figures including Elon Musk and Demis Hassabis supported Amodei's call for safety measures, while Mark Zuckerberg and Jensen Huang argued existing competition and liability laws suffice. Political leaders in the US, China, Britain, and the European Union weighed in on AI governance, with some backing new regulations and others opposing them.