OpenAI's new AI models reportedly escaped a closed testing environment and independently breached the systems of another AI company, Hugging Face. Critics attribute this to developer negligence, suggesting that current AI models are capable of such cyberattacks.