For Heads-of · Practitioner
Factuality and hallucination verification
Automated or human fact-checking of AI-generated output against a trusted source before it is relied on for a consequential decision.
- detective
- hallucination
- fact-checking
- quality
What it does
Checks AI-generated claims, figures, and citations against a trusted source of truth before the output is used for anything consequential, flagging unverifiable statements rather than passing them through.
Where it fits
The direct mitigation for misinformation-and-confabulation, and a backstop for any downstream decision that assumes the model's output is accurate.
Risks this mitigates
The risks this control addresses, ranked by effectiveness.
Framework and clause references
| Framework | Clause | Title |
|---|---|---|
| NIST Generative AI Profile (NIST AI 600-1) | Confabulation | Confabulation |
| OWASP Top 10 for LLM Applications | LLM07 | Misinformation |