AI, creativity and trust: beyond detectors, the question of evidence

Artificial intelligence today draws on vast amounts of content created by humans: books, articles, images, music and videos, often without the consent of their authors. This issue is already a matter of public debate and is beginning to fuel numerous legal disputes around the world, whilst also giving rise to major debates and obstacles concerning training regimes, transparency and copyright infringement.

At the same time, new tools now claim to be able to determine whether a work has been created by a human or by a machine.

This paradox raises a fundamental question: who decides on the authenticity of a work?

Beyond the debates on model training, the protection of creative works or the remuneration of authors, a new issue is emerging: that of trust.

For the reliability of AI detectors remains controversial. Yet their use is becoming increasingly widespread in publishing, education, the media and the creative industries.
Consequently, the debate may no longer pit authors against machines.

Rather, it pits two approaches to trust against one another:

• to rely on detection tools whose results remain a matter of debate;
• to develop provenance and traceability mechanisms that enable the origin and creation process of content to be documented.

The good news is that some of these tools already exist. Content Credentials, provenance mechanisms, the preservation of drafts, revision histories and documentation of the creative process are all ways of building trust in the digital ecosystem.

But their existence alone is not enough.

For businesses, creators, publishers and platforms, the real challenge now lies in their adoption, recognition and evidential value.

FIRSH’s inputs

Having spent centuries protecting works (something we must do even more in the age of AI), we are entering an era in which we will sometimes also need to prove how they were created.

Trust is becoming a legal asset in its own right.

Claire Poirson