Anthropic abruptly revokes OpenAI’s access to Claude API over terms of service violation
Anthropic has revoked OpenAI’s API access to its Claude models after discovering internal use that violated its terms of service. The issue centers on OpenAI staff using Claude Code to support development ahead of GPT-5, which goes against Anthropic’s commercial policy prohibiting the use of its services to build or train competing AI products.
Reports suggest OpenAI integrated Claude directly into internal tools instead of using the standard interface, likely for capability testing and model comparison. While OpenAI acknowledged the restriction, it described such benchmarking as standard industry practice and a spokesperson noted the decision felt one-sided, given that Anthropic can still access OpenAI’s API. Anthropic clarified that OpenAI will still have access for safety evaluations and benchmarking, despite the breach.
The timing is significant, with OpenAI reportedly preparing to launch GPT-5 pretty soon, even as soon as this week, with internal tests already underway.
Comments
The "bullshit generator" wars, the same "scholastic parrot" that FUDDs GNU/Linux, to take down legitimate coverage, with no Quality Check whatsovever... (BS gen is how Dr. Stallman calls LLMs, TL;DR here)
Just stop Gen AI already, it's just a bubble which will never change the world but wreck the planet instead.
Pretty ironic when nor Anthropic nor OpenAI bother about licensing or terms of service when crawling the web. But since Anthropic and OpenAI are (kind of) fighting to get most advanced LLMs (despise exponentially soaring training costs), training one LLM from another have no sense (and would give very bad results), so yes, it's just to benchmark each other. And benchmarks are the main selling point to get more funding.