In the beginning of August, 2025, a major scandal hit the artificial intelligence sphere: Anthropic, a company that develops the Claude AI model, blocked OpenAI access to the Claude API. The decision was made after Anthropic found that OpenAI engineers had used the capabilities of Claude in ways that breached the terms of service set by Anthropic especially in relation to creating the much awaited model GPT-5 model by OpenAI.
Why OpenAI?
Anthropic allowed OpenAI restricted AWS API access to enable them to benchmark Claude with their models, which other AI companies do to warrant safety and quality. But Anthropic accuses OpenAI of more than just a mere benchmarking procedure, saying that it actually engaged in deliberate and methodical tests of Claude Code, which is one of Anthropics most valuable coding assistants, to provide insight, and perhaps a speedier pathway to the development of GPT-5.
Such acts were a breach, Anthropic believes, of commercial terms that expressly forbid the use of the Claude API to:
- Creating or augmenting competing AI products,
- Training competitors models,
- Reverse engineering or copying their services.
According to Anthropic spokesman Christopher Nulty, “Claude Code has been a default coder everywhere choose.” Sadly this is specifically against our guidelines of service.”
OpenAI’s Response
OpenAI argues that benchmarking and testing of their activities are products of the industry, which are crucial in enhancing AI and safety of the community. This was a disappointment and more so as their own API is still open to Anthropic. Anthropic has yet to agree with this transparent access, and cite that this threatens their competitive advantage and the protection of their intellectual property; particularly with GPT-5 said to add advanced coding features; a strong suit that Claude boasts.
Wider Context & Wider Industry
This is not an isolated case, the ability to use and block API access as a means of suppressing competition is becoming a state of the art weapon in the world of technology. This kind of internal fight has already happened in the past like Salesforce closing Slack API access to the competitors, or when Open source platforms like Facebook closed access to competitors to the important data feeds.
Anthropic also recently blocked Claude access to Windsurf, a startup that OpenAI tried to acquire, which furtherment of this trend of blocking tools to competitors. This wider AI ecosystem has treated API access and proprietary model data as a source of competitive advantage and company valuation directly related to the innovation of a company.
What is there to know?
- API access = business weapon and point of contention: Sending or revoking access permissions can have a direct impact on a competitors capacity to innovate or launch new functionality.
- IP and contract compliance issues: With AI companies, terms of service around the APIs is becoming an important issue not merely because it protects the legal rights of those companies but is also a stratagem of the market competition.
- The industry practices are changing: Benchmarking is widespread, but the point at which the fine line between the concepts of an evaluation and creating a competitor is becoming a subject of much debate.
- Ripple effects: These sort of conflicts can result in closer monitoring, more closed ecosystems, that can even impact regulatory oversight.
What’s Next?
With an encroaching threat of AI, as development reaches new speeds and models become even more powerful, get ready to see an even more intense war over access, data and intellectual property. This encounter of Anthropic and OpenAI highlights the shift in AI research social structures that are characterized by collective efforts of the initial field towards the start of the phase of strategic competition, contract enforcement, and access gating.
