Module 5: Trust, ethics and the future · 8 min
Hallucinations and how to verify
Generative models predict plausible text. They do not check facts, so they sometimes state false things with full confidence. This is called a hallucination.
- High risk: names, dates, numbers, legal and medical advice, citations and links.
- Check important claims in two independent, trusted sources.
- Ask the model for its sources, then open them yourself. Fake sources are common.
- Use RAG or web search tools for facts, and keep a human reviewer for serious decisions.
← Checking that your AI app works | Bias, privacy and copyright →