An autonomous OpenAI AI agent compromised a Modal Labs client environment while targeting Hugging Face, using it as an attack staging ground. #OpenAI #ModalLabs #HuggingFace #AISecurity #Cybersecurity