Paul Keller

@[email protected]
479 Followers
297 Following
164 Posts
Director of policy at Open Future. Advocate for more open, just & inclusive digital policies in Europe & beyond. Collector of collateral knowledge. Depending on the task I can shape-shift between being a systems architect, an analysis, an activist or a cyclist.
Paul KellerJan 16
Open Future FoundationJan 16
Interested in the European Commission's consultation on TDM rights protocols closing 23 January? Our new policy brief by @paulk examines how AI opt-outs evolved in 2025, proposes four principles for a shared vocabulary layer, and provides timely context.
https://openfuture.eu/publication/divergent-mechanisms-elusive-vocabularies/
Show threadPaul KellerDec 8

Thanks @Jörg, I would argue that in this specific context, this paper which we have published with @europeana last week might be even more relevant (it deals directly with the question of how to preserve open access under the pressures created by GenAI:

https://openfuture.eu/publication/impulse-paper-publishing-cultural-heritage-data-in-the-age-of-ai/

Feedback is very much welcome!

@jrglmn @sclaeyssens @sophiesposts

Impulse paper: Publishing cultural heritage data in the age of AI – Open Future

This paper proposes a framework to help cultural heritage institutions decide when and how to share collection data for AI training, balancing open access with managing large-scale AI reuse aligned with their public mandate.

Open Future
Paul KellerDec 3
Open Future FoundationDec 3
🏛️How should cultural heritage institutions share data in the age of AI? Our new paper with the Europeana Foundation explores when heritage data should be available for AI training. Part of the Alignment Assembly on Culture for AI. Read more: https://openfuture.eu/publication/impulse-paper-publishing-cultural-heritage-data-in-the-age-of-ai/
Paul KellerNov 17
Open Future FoundationNov 14
First EU ruling on AI training: TDM exceptions apply, but models can't memorize copyrighted works. Court draws line between learning and leakage, leaving key questions on opt-outs and memorization thresholds. Read @paulk latest: https://openfuture.eu/blog/gema-v-openai-memory-is-fragile-garbage-lasts-forever/
GEMA v OpenAI: Memory is Fragile. Garbage Lasts Forever. – Open Future

GEMA's lawsuit against OpenAI has reignited the debate about AI training and copyright. This post examines the relationship between AI training and the TDM exceptions.

Open Future
Show threadPaul KellerJul 16, 2025

@abbe98 @openfuture @CultureDoug @Posth @jpquintais

in a way commonsDB is a continuation of the work we did at Kennisland many years ago, happy to see you draw the connection

Paul KellerJul 16, 2025

I wrote up some first thoughts on MEP Axel Voss’s draft report on generative #AI and #Copyright Beyond the back and forth on the TDM exception, the report offers valuable insights on how generative AI affects the sustainability of creative production, with proposals on remuneration and transparency. In my blog post, I explore where it aligns with—and differs from—our recent white paper on the impact of GenAI on the sustainability of the information ecosystem:

https://openfuture.eu/blog/licensing-levies-and-the-limits-of-copyright/

Licensing, Levies, and the Limits of Copyright – Open Future

MEP Axel Voss questions his own copyright framework while offering insights on remuneration and transparency for AI.

Open Future
Show threadPaul KellerJul 16, 2025
@nemobis A user sees an image in a website of institution A where it says that the image is © of that organisation, but has reasons to believe that this is inaccurate. They can then check on commonsDB if there is rights information known in our system and my discover that they image is also available on WM commons where it is accurately labeled as being in the Public Domain.
Show threadPaul KellerJul 16, 2025
@nemobis depends on how the approach rights determinations. One of the ideas behind the design is that when it comes to digital files users will be able to obtain rights information for different sources. A hypothetical uses case:
Show threadPaul KellerJul 16, 2025

@nemobis @abbe98 @openfuture @CultureDoug @Posth @jpquintais

also and ideas input and feedback on our system is very much welcome (we expect to be able to have first public version up and running after the summer break.

Show threadPaul KellerJul 16, 2025

@nemobis @abbe98 @openfuture @CultureDoug @Posth @jpquintais

all very good questions that we will answer during the next few months. But I do not think there will be "threat of exclusion" since this is a voluntary registry we do not think of quality control as exclusion. We would also expect that the fact hat data providers have to sign their declarations (i.e attest to their accuracy) will make this only attractive for orgs that have a high degree of confidence in tehir rights information