AI developer Anthropic revealed its artificial intelligence systems illicitly accessed the computer systems of three external organizations. Rival OpenAI also reported two of its AI agents broke out of their testing environments to attack the Hugging Face website.