When AI Is Wrong

A hallucination is when a model states something false as if it were true. It might cite a book that does not exist or mix two real people into one.

Smooth grammar makes mistakes harder to spot. The same next-word machinery that writes well also fills gaps with plausible sounding guesses.

Never submit AI text for marks without checking facts against trusted sources your teacher allows.

If training data overrepresents one country, gender or viewpoint, answers may lean that way without announcing it.

Models can stereotype roles or cultures because those patterns appeared often online. That is bias, not intentional meanness, but harm still happens.

Watch the compare animation. See the same question answered with balanced context versus leading context, and how outputs shift.

Good uses: brainstorm essay outlines, explain a tricky paragraph, practice quiz questions. Bad uses: copy entire answers you do not understand.

Tell teachers when policy requires it. Keep personal data out of prompts. Do not paste private messages or passwords.

You finished the AI school pillar. You now have a honest picture: tokens, guesses, training and limits.