1. AI-assisted evaluation
Scores, issue detection, summaries, recommendations, comparisons, and safety verdicts are generated with the help of AI models.
These results are intended as decision-support signals, not final judgments.
2. No guarantee of accuracy
EvalLoop AI does not guarantee that an evaluation is correct, complete, unbiased, or suitable for your specific use case.
AI models can make mistakes, misinterpret context, or produce inconsistent results.
3. Human review is important
Always review important AI outputs manually, especially for:
- Medical or health-related content
- Legal or financial decisions
- Safety-critical instructions
- Personal, confidential, or sensitive information
- Published customer-facing content
- High-risk business decisions
4. Safety checks
AI Safety checks are designed to highlight potential risks, but they cannot detect every possible safety issue.
A “Safe” result does not guarantee that content is safe. A “Needs Review” or “Unsafe” result should be reviewed carefully by a person.
5. Sensitive data
Do not submit passwords, private keys, personal identity documents, financial details, confidential business information, or other sensitive data.
6. Beta notice
EvalLoop AI is in beta. Features, evaluation methods, models, limits, and results may change.
7. Your responsibility
You are responsible for how you use evaluation results and for any decisions you make based on them.