How Much Truth Can a Society Handle in the Age of Deepfakes?

I see deepfakes as a very serious threat. When you also consider that a great deal of politics seems to be conducted via platforms like X, a breeding ground for false statements, hate, incitement, and fakes, it can make you feel uneasy very quickly.

But it doesn't even have to be about politics. Many people have already experienced it: The phone rings, and then you hear the voice of your own daughter, allegedly in trouble and urgently needing money .... far too many people have already fallen for it.

These times are definitely over: “If it looks like a duck, quacks like a duck, and walks like a duck, then it’s a duck.”

The duck test has unfortunately outlived its usefulness — at least as evidence. In the future, something can look, quack, and walk like a duck and still be nothing more than a very well-programmed duck remix.

What seems almost even more dangerous to me than the individual fake is the blanket claim that “anything can be faked.” That is precisely how politicians, companies, and private individuals can evade responsibility. A genuine recording can then simply be labeled an AI fake until no one knows what they can rely on anymore. This possibility even has a name, I believe: the so-called liar’s dividend.

In everyday life, it is therefore likely that a new habit will help more than perfect technical detection: rather than reacting to the dramatic message, call back via a second channel first, look for another source, and buy some time. In the case of the supposed daughter on the phone, agreeing on a code word in advance might be useful — it sounds a bit like something out of a spy movie, but it is probably cheaper than the entire deepfake defense industry.

For political content, there would also need to be verifiable provenance records for media, rapid corrections, and clear rules for platforms. But even that will not be a truth stamp. In the end, media literacy will probably mean being more willing to tolerate uncertainty: “I don’t know yet” may then be the most sensible answer, even if, by nature, it gets the fewest likes online.

…but they are not a magical seal of truth. A signed recording initially says only that it came from a specific device or account. The device may have been hacked, the context may be missing, and a genuine recording can still be misleadingly edited. Unfortunately, the truth rarely comes with a pretty seal of approval.

That is why a multistage process would be more important: the file’s origin, the time and place, independent confirmations, and transparent documentation of the review. That should be standard practice in newsrooms and government agencies. In everyday private life, one can adopt this simple habit: don’t forward something immediately, use a second way of contacting the person, and always pause when money is being requested. If the supposed daughter is crying on the phone, “I’ll call you back on the number you have saved” is not distrust but common sense.

I find it problematic when platforms merely display warnings and thereby shift the responsibility onto individual users. A small notice saying “possibly manipulated” is of little help if the post has already been shared millions of times. At the same time, I do not want a central truth authority that decides, according to opaque rules, what is allowed to remain visible. Perhaps what is needed instead are independent review bodies with publicly disclosed criteria and clear avenues for appeal.

In the end, the decisive resource is probably not perfect certainty but trust in procedures. In other words: Who reviewed it, according to what rules, and what remains unclear? That is less spectacular than a “genuine” video, but presumably a more robust form of truth. The question then would be whether we as a society are willing to endure such slow procedures when the digital pub talk has long since launched its next round of outrage.