Research Note
Conversation Example Test Specification
Each example should identify the user situation, input, relevant policy or source, expected answer properties, prohibited behavior, evaluation method, and consequence of
Conversation Example Test Specification
Each example should identify the user situation, input, relevant policy or source, expected answer properties, prohibited behavior, evaluation method, and consequence of failure.
Examples should cover routine work, ambiguity, missing information, conflict, out-of-scope requests, unsafe requests, adversarial phrasing, sensitive data, and recovery after a bad turn.
Examples in configuration can influence behavior. A separate blind test set is needed to evaluate whether the behavior generalizes beyond memorized patterns.
Sources
Follow the evidence.
- owasp.org: www project top 10 for large language model applicationsowasp.org
- open.spotify.com: 3keuOAMwBSyXBXpxmisUr6open.spotify.com
- genai.owasp.org: llm01 prompt injectiongenai.owasp.org
- about.fb.com: create your own custom ai with ai studioabout.fb.com
- NIST AI Risk Management Frameworknist.gov
- genai.owasp.org: owasp top 10 for llm applications 2025genai.owasp.org
- ai.google.dev: prompting strategiesai.google.dev
- privacycenter.instagram.com: policyprivacycenter.instagram.com
- Introducing the Meta AI appabout.fb.com
- about.fb.com: metas approach to labeling ai generated content and manipulated mediaabout.fb.com
- daltonanderson.ghost.io: build your ai agent with meta ai studio no code neededdaltonanderson.ghost.io
- genai.owasp.org: announcing the owasp gen ai red teaming guidegenai.owasp.org
- ai.meta.com: ai studioai.meta.com
- facebook.com: policyfacebook.com
- NIST: Artificial Intelligence Risk Management Framework, Generative Artificial Intelligence Profilenist.gov
- youtu.be: zlpebV6cHYyoutu.be
- Meta generative AI privacy guidefacebook.com