Anthropic bans OpenAI's access to Claude due to ToS violation

The battle between companies focusing on AI endeavors is heating up and OpenAI has found itself in rather unfavorable positions in recent times. First, many of its employees were poached by Meta, despite efforts to deter staff from accepting Mark Zuckerberg's "ridiculous" offers. Now, it has been banned from using Anthropic's Claude following a violation of the terms of service (ToS).

Multiple sources confirmed to Wired that Anthropic had cut off OpenAI's access to the Claude API due to an apparent violation of ToS. An Anthropic spokesperson confirmed to the outlet that the banhammer struck earlier this week after Anthropic discovered that OpenAI's technical staff were using Claude in their development of GPT-5. While this wouldn't sound like a problem since Claude is meant to help developers with coding, it is a violation of Anthropic's ToS which restricts people from using Claude to build commercial competing products, including competing AI models.

Sources have also revealed that OpenAI was allegedly plugging Claude's API in their internal development tools rather than using the chat interface. As such, it was also evaluating Claude against various inputs and metrics and making adjustments to its own AI models accordingly so that it performs better than Claude under similar circumstances. OpenAI admitted that it was benchmarking Claude, but downplayed the charge by emphasizing that this is industry-standard.

Anthropic did agree that benchmarking AI models is industry-standard and that OpenAI could have access to the API specifically for this purpose, it didn't go into details about how it would enable this activity given the existing ban.

This isn't Anthropic's first rodeo with a competing AI company. In June, it suddenly cut off Windsurf's access to Claude and while it didn't specify a reason publicly, there is strong suspicion that it was due to rumors of the IDE being acquired by OpenAI; that deal fell through though.