Deepfaked media

Deepfake is the term used to describe the use of artificial intelligence techniques that alter digital media, mainly video, audio or images. As we have seen in recent months, its most popular application is swapping faces and voices between people.

Tools for creating deepfakes are becoming ever more accessible. And although, in principle, they are designed for harmless fun and entertainment, the synthetic content they enable could be misused to manipulate viewers and sow distrust.

Using deep learning algorithms, deepfakes are already impossible to spot by eye and ear. This means that only detection technology can identify advanced forms of manipulated and altered media.

More and more apps use neural networks to analyse photos. The AI learns from its mistakes through training based on deconstructing and reconstructing faces. Although it may look like fun at first sight, the danger comes when this technology is used to defame, persuade or polarise debate.

We have already seen how the quality of these synthetic media has reached the point where even news outlets struggle to tell real news from fake and to detect these manipulations, which erode trust: who is asserting the lie? Who is denying the truth?

There is a race to develop tools that can detect deepfakes, with the aim of defending the truth.