ray-train

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The HuggingFace example calls AutoModelForCausalLM.from_pretrained(...) and AutoTokenizer.from_pretrained(...), which will fetch user-contributed models/tokenizers from the public HuggingFace model hub (an untrusted third‑party source) as part of the runtime workflow.