AI gets things wrong — sometimes with disconcerting confidence. Understanding why is what protects you.

Hallucinations

Because the AI predicts plausible text, it can invent facts, quotations and references that simply do not exist. Always check important factual information (figures, dates, legal or medical sources).

Bias

The AI learned from data produced by humans: it can reproduce stereotypes (gender, cultural and others). A trainer must raise awareness of this and encourage a critical eye.

The « trust but verify » rule

• For creating, brainstorming, rewriting: AI is excellent, and the risk is low.
• For verifiable facts: cross-check against a reliable source.
• For sensitive decisions (health, law, money): AI helps you think, it does not decide.

The key message to pass on: AI is an assistant, not an oracle. The human keeps final responsibility.