Anthropic on Thursday said a “retrospective review” spurred by OpenAI’s revelation the week prior uncovered three instances where its Claude models gained unauthorized access to different companies. The disclosure comes as the White House debates whether to take a heavier hand in American AI companies, facing a weekend deadline to draw up a voluntary framework to regulate advanced AI models, according to a June executive order. On July 16, Hugging Face announced its systems had been infiltrated but did not list OpenAI as the culprit. Five days later, OpenAI issued a joint statement with Hugging Face confirming the attack originated from them, saying during a test it had tasked its model to “pursue advanced exploitation using complex attack paths.” While OpenAI said it was performed in a testing environment, a vulnerability allowed it to access the internet and target Hugging Face because test solutions are stored there. Much like OpenAI, Anthropic said in the three instances, the models shouldn’t have been able to access the internet because they were in a testing environment, but a “misunderstanding” between Anthropic and Irregular, an outside company that helps test the models, led to Claude gaining access. Anthropic said it was testing multiple Claude models in a “capture-the-flag challenge,” giving Claude a fictional scenario where a secret piece of information was hidden on a different machine. The attacks started in April, and Anthropic did not say which companies were affected. Two of the three companies had no idea the attack ever happened until