aws-ai-ml

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

• Third-party content exposure detected (medium risk: 0.30). In references/dataset-evaluation/overview.md, the runtime workflow accepts a user-provided/local or S3 file path and passes the resolved dataset file contents to scripts/format_detector.py for JSONL format validation, which means outsider-authored dataset text is ingested into the LLM runtime via the user’s file.