Mark học AI

Video #122 · The dark side of AI · Part 5/5

How to spot AI-made images: watermarks, C2PA and fact-checking

A breaking-news photo is all over the internet: is it real or made by AI?

Watch videoVideo in Vietnamese

Real photo or AI photo?

Mark sees a breaking-news photo spreading everywhere online, and can't tell whether it's real or made by AI.

WATERMARK

A hidden mark in the image

Bit explains: many AI tools now add a watermark, a hidden mark you can't see with your eyes but a scanner can read.

C2PA

A birth certificate for photos

C2PA is a standard that attaches origin information to the file, like a birth certificate for the photo: what made it and when it was edited. On some sites, you can click the Content Credentials icon to see it.

AI LABELS

Labels on social media

Big social networks also label AI-made content, and Europe and many other countries, including Vietnam, are writing rules for marking AI content into law.

LIMITS

No label doesn't mean real

But the mark can be lost when the image is cropped, screenshotted or reposted. No label doesn't mean the photo is real.

CHECK IT

Three steps to check

For important news, do three things: run a reverse image search, see who posted it first, and compare with mainstream news outlets.

This article is based on the video Spotting AI-made content (watermark, C2PA) from the Mark học AI channel. Watch the video (in Vietnamese) to see the animations.