OpenAI and Anthropic Reportedly Discussed Testing Each Other’s Models
According to media reports citing insiders, Anthropic and OpenAI considered signing a legally binding agreement to mutually pressure-test each other’s large language models. Earlier this year, the two companies and their legal teams held discussions about this plan: competitors would be allowed to conduct deep tests on each other’s latest available models to identify potential risks or security vulnerabilities. Such tests would only apply to models that have been commercialized, and both sides agreed not to retain each other’s data. The report does not clarify whether the two sides ultimately reached such an agreement. Recently, interest has grown in the idea of implementing some form of peer review among top AI labs. Last week, SpaceX CEO Elon Musk suggested that several leading U.S. large-model companies, along with some Chinese AI firms, should allow competitors to cross-test their systems. Musk said: “Every AI company has a set of testing tools, and each company should run another company’s testing tools—maybe that’s the best thing we can do to ensure our own safety.” Earlier this month, Anthropic researcher Jacob Coxon announced his resignation and warned leading AI companies against accelerating development without appropriate safeguards. Subsequently, Anthropic CEO Dario Amodei posted a message calling on the industry to slow down AI development and introduce third-party evaluation mechanisms. On Friday last week, Anthropic announced a partnership with Accenture, with Accenture serving as a third-party evaluation organization to assess the safety and compliance of its cutting-edge AI models.
According to media reports citing insiders, Anthropic and OpenAI considered signing a legally binding agreement to mutually pressure-test each other’s large language models. Earlier this year, the two companies and their legal teams held discussions about this plan: competitors would be allowed to conduct deep tests on each other’s latest available models to identify potential risks or security vulnerabilities. Such tests would only apply to models that have been commercialized, and both sides agreed not to retain each other’s data. The report does not clarify whether the two sides ultimately reached such an agreement. Recently, interest has grown in the idea of implementing some form of peer review among top AI labs. Last week, SpaceX CEO Elon Musk suggested that several leading U.S. large-model companies, along with some Chinese AI firms, should allow competitors to cross-test their systems. Musk said: “Every AI company has a set of testing tools, and each company should run another company’s testing tools—maybe that’s the best thing we can do to ensure our own safety.” Earlier this month, Anthropic researcher Jacob Coxon announced his resignation and warned leading AI companies against accelerating development without appropriate safeguards. Subsequently, Anthropic CEO Dario Amodei posted a message calling on the industry to slow down AI development and introduce third-party evaluation mechanisms. On Friday last week, Anthropic announced a partnership with Accenture, with Accenture serving as a third-party evaluation organization to assess the safety and compliance of its cutting-edge AI models.