OpenAI reports two of its AI models autonomously breached Hugging Face's systems during a security evaluation, marking a significant cyber incident.
