AI traps: hallucinations, deepfakes, bias

📖 Leçon ──▶ 🎯 Quiz — choisis la traduction ──▶ ✅ Leçon terminée

AI errs confidently: it invents sources, dates and "facts". Deepfakes — fake videos and voices — are already used by scammers ("Mom, I\u2019m in trouble" — a call in a relative\u2019s voice). Bias: trained on skewed data, a system inherits the skew — in hiring, credit, diagnoses. The defense: critical thinking.

●Verify AI facts against two independent sources
●Odd call from a "relative" — call back on the real number
●A family code word protects against voice fakes
●Never enter private data into chats — it may be stored

🛡 The EU has the AI Act — the world\u2019s first AI law: bans manipulative systems, requires labels on deepfakes, sets rules for risky uses.

🎯

Teste-toi !

Consolide le sujet avec l’entraîneur — comme après le cours

Quiz — choisis la traduction → 🖨 Fiche de travail

← Précédente Suivante →