Staying informed when everything can be fabricated

A photo, a voice, a video can now be produced in minutes. The consequence is not the one you might expect, and that is the subject of this lesson.

The real danger is not the fake

We imagine the risk like this: we will be shown false things and we will believe them. That is real, but it is the least serious part.

The main danger is the opposite. When everything can be fabricated, then everything can be denied. An authentic and compromising video becomes contestable with a single phrase: "it's AI". This doubt requires no proof; it is enough that it be plausible.

What is being put in place

The main technical response is called provenance: embedding in the file, from the moment of its creation, where it came from and how it has been modified. Invisible watermarks are beginning to be made mandatory by law in some territories, and a common standard is being rolled out among device manufacturers and software vendors.

It is useful, and it is insufficient: these marks can be removed, a screenshot is often enough to erase them, and content without a mark is not necessarily fake. Provenance proves origin when it is present; its absence proves nothing.

The habits that actually work

1. Look for the source, not the clues. Hunting for malformed fingers is a skill that is outdated in advance. Instead, ask: who is publishing this, and where was it before?
2. Go back one step. An image shared by a friend came from somewhere. Does that somewhere exist, and does it say the same thing?
3. Cross-check on a different medium. If an event took place, it does not exist in a single video alone.
4. Beware of what suits you. A fake that confirms what you already thought gets past every defence.

And everyday information

Another change, less spectacular and more profound: search engines now answer directly, instead of sending users to websites. For you, that is faster. For those who produce the information, it is a source of traffic that disappears, and therefore of means to investigate.

There is no simple right answer in all this. But there is one gesture: when an answer has served you well, open the source. It is the only link left between what you read and those who produced it.

To do right away: the next time an AI answer cites an article to you, open it. You will see along the way whether it actually read it.