Hugging Face Inference Infrastructure Compromise via Malicious Model with `trust_remote_code=True` | AI Breaking Wire