AI biases 🟢
An AI is not “neutral”. It learns from data made by humans — and it inherits their errors and biases.
sometimes unbalanced
biases too
which seems “objective”
Example
If an AI has only seen photos of male doctors, it will “think” that a doctor must be a man. This is not malice — it is a reflection of data .
Disinformation & deepfakes
A deepfake is a video, an image or a voice manufactured by AI to make someone say or do something… who has never done it.
🔍 How to spot them
The eyes
Fixed gaze, odd or absent blinks.
The mouth
Lips poorly synchronized with sound.
The light
Shadows that don't stick to the face.
hands
Extra, deformed or fuzzy fingers.
The voice
Your flat, absent breathing, robotic.
The source
Where does the video come from? Who published it?
Check information: the checklist
Before believing (and sharing), filter the information. Click to check each reflex.
- Who said it? (known source, serious?)
- Do other reliable sources say this too?
- When is the information from? (not an old photo released)
- Is this trying to make me react very strongly? ⚠
- Is this proof, or just an opinion/image?
- Did I check on a fact-checking site?
Privacy & personal data
AIs love data. What you give them can be stored, read, reused. Protect what is yours.
To do
- Think before pasting personal text into an AI
- Use a nickname when possible
- Read (at least quickly) what we accept
- Ask an adult if in doubt
To avoid
- Give your address, your school, your number
- Send private photos to an AI
- Share others' secrets
- Believing that “it’s private” by default