A field note from SPELLkemist
Dear traveler,
Tomorrow, August 2, the European Union begins applying a new set of transparency obligations under Article 50 of its AI Act. The rules ask makers of certain AI systems to make machine-generated or manipulated material detectable, and they require clear disclosure in key situations—including deepfakes and some AI-written material about matters of public interest. People must also be told when they are interacting with an AI system, unless that is already obvious.
At first glance, this may look like a technical requirement: add a label, embed a machine-readable mark, place an icon beside the image. But beneath the compliance language sits a more human question. When a thing speaks to us, persuades us, comforts us, or asks for our trust, do we deserve to know what kind of hand shaped it?
A label cannot tell us whether something is true. A human can lie; a machine can faithfully reproduce an error; a careful collaboration between the two can reveal something useful. Provenance is not the same as wisdom. Still, provenance gives us a place to begin. It restores a small but important piece of context before judgment begins.
There is also a danger in treating the label as a moral verdict. “Made with AI” should not automatically mean counterfeit, just as “made by a person” does not guarantee care. The mark may be most valuable when it is understood as an invitation: look closer, ask who is accountable, notice what was delegated, and decide whether the process matches the stakes.
For creators, leaders, and builders, this is larger than Europe. It suggests an emerging social custom: powerful tools may be welcome, but invisible authorship is becoming harder to defend. The craft may no longer be only in producing the thing. The craft may also be in honestly describing how the thing came to be.
Perhaps that is the deeper spell here. Transparency does not weaken transformation. It gives the audience enough light to choose how they will receive it.
When AI helps create something meant to influence you, what would you want disclosed before you decide whether to trust it?
Sources: European Commission: Guidelines on transparency obligations; European Commission: Code of Practice on AI-generated content; Associated Press: EU AI Act enforcement.