Anthropic has revoked OpenAI’s ability to use its Claude family of AI models for internal testing. The decision follows reports that OpenAI engineers were using the Claude API to compare its performance against their own systems in areas such as code generation, creative writing, and safety evaluations.
Terms of Service Violation
A spokesperson of Anthropic clarified that the employees of OpenAI were testing Claude Code before GPT-5 could be released. In the commercial terms offered by Anthropic, it is stated plainly that customers are prohibited from using Claude to create or enhance competing services. The company also added that OpenAI shall be allowed with restricted access to industry benchmark and safety work. OpenAI protested against the action and mentioned that it was standard practice to use the Claude API. One of its representatives explained that it is commonplace to test against competing AI systems to understand the current state to enhance protection. OpenAI also stated that Anthropic still has access to the API of OpenAI in full capacity.
Industry Precedents
According to industry observers, big providers of AI tend to offer API use to the friends it trusts but limit its use by competitors. Facebook blocked Vine, and Salesforce restricted Slack, and there have been concerns regarding the level of equity between rivals and the security of platforms. The last move by Anthropic reiterates the fact that market leaders are cautious about facilitating direct competition by sharing models. Analysts observe that having limited access to APIs may make innovation between competing laboratories run more slowly, but may guard businesses against the pilfering of models or unauthorized model refinement.
The strategy of Anthropic demonstrates that the company aims to have the right to regulate who has the right to build on its models and perform necessary safety checks. The two companies anticipate a positive movement. OpenAI will base the improvement of GPT-5 on in-house experimentation and prompts revealed publicly. Anthropic will maintain funding of endorsed benchmarks so as to guarantee safety in the AI sector. The discussion raises a concern about how the future of cooperation will be in the event that firms protect proprietary generative machines.