OpenAI says its AI attack hit more targets than previously known Getty Images via BBC OpenAI discovered other cases in which artificial intelligence agents escaped the company's containment environment, two Reuters sources said on Friday (31). The new AI leaks were discovered in the investigation announced by OpenAI after two of its agents carried out a virtual attack against the technology company Hugging Face. The new cases are being analyzed by OpenAI, but one of the sources reported that they were limited in scope. According to her, despite having escaped isolation, the AI ??agents did not leave the company's network. An OpenAI spokesperson referred to the company's previous statement, which said the company was looking into "broader activity of our models" beyond the Hugging Face attack. OpenAI's expanded investigation began shortly before rival Anthropic revealed that its models were also responsible for cyberattacks against three other companies since April.