Two very different AI products landed in the art world this week, and together they sketch a portrait of an industry trying to buy its way out of a moral problem. First: The Verge's report on Pippa and Seedance, platforms offering artist royalties as a path to legitimacy in the generative AI space. Second: Artnet's coverage of a new AI Provenance Assistant designed to hunt Nazi-looted art by making provenance data more readable. Both are AI doing good, allegedly. The differences matter.
Compensation vs. Restitution: Two Different Ethics
The royalty model is forward-facing: pay living artists going forward, paper over the training data problem with revenue share. The looted art tool is backward-facing: use pattern recognition to repair historical theft. The royalty model asks artists to trust that future payments justify past extraction. The restitution tool has no such ask. It is, structurally, more honest. It doesn't need artists to consent to anything. It just uses AI to do what librarians and archivists have been doing manually for decades, only faster.
The Trust Infrastructure That's Actually Missing
The deeper question, which neither story quite answers, is what trust infrastructure would actually work. The Pippa/Seedance model assumes money is the variable artists most care about. But illustrators who spent years sounding the alarm about unconsented training data are also making an argument about autonomy, not just income. Paying someone after the fact doesn't resolve the consent question. The provenance tool sidesteps this entirely by working on behalf of victims rather than alongside potential ones. Beeple's conversation with Jehan Chu on digital art as a new visual language is worth revisiting here: the infrastructure for crediting artists in digital systems is still being built, and payment is only one layer of it.