Did Openai use the coding tool of rival cloud to train GPT 5? Anthropic says this is possible
Anthropic claims that OpenIAI has used Cloud’s coding tools to train its next-gene model, GPT5.
Listen to the story
In short
- Rules cut the access to anthropic openi for the cloud model on the breech
- According to claims, Openai used anthropic API for deep testing of GPT5
- Anthropic forbids using its model to create competitive AI system as a rule
The Battle of AI Titans has taken a new turn. Since Openai is ready to launch its next-gene AI model, the anthropic claim GPT 5 is based on its cloud model. Anthropic has recently announced that it is officially cutting the access to the openIA to its cloud model, accusing the chat of breaking the rules to give itself an edge. The story reported by Wired for the first time claims that openiI engineers were using the cloud of Anthropic, especially its coding tools, as part of the internal test to the benchmark and refine their own models.
Anthropic spokesperson Christopher Nalty told Wired, “Cloud code has become a goto choice for Kodar everywhere, and so it was no surprise to know that OpenaiI’s technical staff were also using our coding tool before the launch of GPT 5.” Unfortunately for Openai, it was clearly a step far away. “Unfortunately, this is a direct violation of the terms of our service,” said the copycat.
What has happened here if we go through anthropic claims. Instead of using the cloud through the general chat interface like regular users, Openai allegedly plugged into the developer API of Anthropic directly. This allowed the company to run deep tests of cloud capabilities in areas such as coding, creative writing, and even sensitive signals about safety issues including CSAM, Selfahm and defamation.
The idea was how the cloud behaves, compare the results with your model of openaiAI, and make adjustments before rolling your next generation system.
Meanwhile, the anthropic day 1 was very clear on your commercial terms, you can not use its model “to manufacture a competitive product or service, including training the competing AI model or removing the service.”
According to anthropic, it was exactly opening. However, the tube made it clear that the anthropic would not completely cut Openai. He said, “We will continue to ensure that OpenIA has API access to the purposes of benchmarking and safety evaluation, as is the standard exercise in the industry,” he said.
Openai is not thrilled about blocks, but they are eager to keep things citizens. Chief Communications Officer Hannah Wong issued a statement for Wired, “This is an industry standard to evaluate other AI systems to evaluate other AI systems and improve safety. While we respect anthropic’s decision to cut our API access, it is disappointing to be available for our APIs.”
Interestingly, this is not the first time anthropic has shown a rival door. Earlier this year, the company allegedly blocked a codingfocked AI startup windsurf, with rumors to reach the cloud after the spread of rumors that OpenaiI was proceeding to achieve it.
Subsequently, the Chief Science Officer of Anthropic, Jered Coupon dedicated the situation: “I think it would be strange for us to sell Cloud to Openai.”
This latest episode shows how competitive the AI race has become. On the horizon with GPT 5, every model, and every line of the code matters.
