Discussion
Under what conditions should copyright-protected content be used for AI training?
🗣️We know this is a topic many of you feel strongly about.
We want to hear your views on issues related to the use of copyright-protected content for training AI models, for research, and more.
📝Consultation open until 3 November 2026.
@EUCommission sollte nur erlaubt sein, wenn LLMs vergesellschaftet sind, d.h. unter öffentlicher Kontrolle und ohne Gewinnabsichten.
Abgesehen davon: die real existierenden LLMs haben bereits alles Wissen reingesogen, was es so gibt auf der Welt. Mit eurer Fragestellung seid ihr um Jahre zu spät.
@EUCommission if the bot is trained on copyright, it can pay royalties for every output it generates
Quatsch , nur auf Englisch
Vielleicht sollte die Kommission zuerst ihre eigenen Regeln einhalten.
Öffentliche Konsultationen
Um den Bürgerinnen und Bürgern die Teilhabe an der EU-Politik und an Rechtsetzungsverfahren zu erleichtern, veröffentlicht die Europäische Kommission regelmäßig öffentliche Konsultationen in Form von Online-Fragebögen. Diese Fragebögen werden IMMER mindestens ins Englische, Französische und Deutsche übersetzt …
I see a "slight" mismatch.
Why should AI companies be able to train their models with copyright-protected content for free, while people have to pay to train their minds with copyright-protected content? If copyright laws apply, both should pay. If they don't apply, neither of them should pay.
@EUCommission Under no circumstance without license.
La commission commence à nous les briser menu en ne respectant pas ses propres règles
Consultations publiques
La Commission organise régulièrement des consultations publiques sous la forme de questionnaires en ligne, afin de faciliter la participation des citoyens au processus politique et législatif de l’UE. Ces questionnaires sont TOUJOURS disponibles au moins en anglais, français et allemand, et souvent dans d’autres langues officielles de l’UE.
@EUCommission simply apply the basic copyright maybe ?
Using copyrighted material = illegal.
@EUCommission I am not sure about this. Every person who ever lived learned what they know from someone else's work. Oppenheimer-Einstein-Newton etc.. If we build a tool, shouldnt it be the best tool that it can be?
@EUCommission Suppose I want to research a video compression algorithm. Should I be allowed to download copyrighted movies, process them, and view the result in order to evaluate the effectiveness of my compression? The law is, I can't do that without explicit approval from the copyright owners, no matter what benefits I might claim to hope to obtain from my research.
AI training is not fundamentally different.
@jbqueru @EUCommission if you want to make a legal analogy i think this one is slightly flawed. evaluation of effectiveness does not tend to produce a derived work in the manner that is typically under consideration in copyright law. in my discussion of compression libraries at https://codeberg.org/cosmicexplorer/corporeal i specifically refer to the publication of compression dictionaries as an example of a derived work not allowed by (but expressly covered by) copyright law (i.e. an infringing publication).
@hipsterelectron
That's fair.
The case of compression dictionaries is weird, though, because that depends on the nature and size of the dictionary.
Statistics about the frequencies of individual letters indeed are probably OK. Lists of words in alphabetical order are probably OK.
But a list of the 100 billion most common sentences in the English language, along with statistics about which sentence might follow which, not so immediately clear.
@jbqueru @EUCommission this is why i am a strong advocate of the US-centric policy of fair use which effectively adjudicates this through multi-factor balancing tests. if the EU wants to adopt that policy they can be my guest
@hipsterelectron Interestingly, I really don't like US fair use. There's no actual law other than "see you in court", which in turn means that the big guys win over the small ones. Even if a small guy somehow prevails, they're still drowned under the cost of the legal process.
@jbqueru @EUCommission completely disagree and think this is a poor argument
@wyatt i'm not contesting that, i'm contesting that being remotely related to fair use
@jbqueru @EUCommission the EU and most countries are much stricter than the US and have no concept of transformative fair use, but even in the US the act of producing generative model output lands squarely in the range of overlapping with the secondary market for the original work, which was found to be unlawful in hachette vs IA ("internet archive") twice in a row recently. so in every possible case this is obviously against copyright law and every "just asking questions" post from the EU is abject bullshit.
@jbqueru @EUCommission it's cool that the EU is taking opinion polls to change legal standards established across centuries though. maybe we should take an opinion poll about border control and see how many people think frontex should exist
@hipsterelectron @jbqueru @EUCommission
Surprisingly many, cause the opinion they shouldn't exist at all is kinda not that widespread.
However simultaneously also a majority thinks border controls INSIDE the EU shouldn't exist and need to be abolished yesterday...
So nobody apparently gets what they want...
@agowa338 I certainly appreciate that @EUCommission is reaching out through different channels. The experience with CRA was scary, where individual creators were initially ignored until there was a massive "Oops, we forgot all the people who don't have lobbyists and legal departments", so I appreciate that, at least, there's an effort to cross-check whether the sentiments expressed on social media have a counterpart in the official work.
@EUCommission Why care to ask if at the end you (Ursula and friends) will do whatever technooligarchs will tell you to do
It's about a money and power .
Useless UE.
Mal vamos, si para empezar sólo se va a permitir contribuciones en inglés, francés o alemán. Estaba convencido de que la Unión Europea estaba compuesta por más paises.
None. There are NO circumstances in which copyright law does not apply to AI and the companies & individuals who let it loose on the world.
Make them pay. Make it prohibitively expensive so they don’t just keep stealing.
Yes, stealing.
@EUCommission@ec.social-network.europa.eu Under the conditions of fair and impartial negotiations with the copyright holders, during which all involved parties (no proxies) have come to a mutually satisfactory agreement.
The legislation already exists; politicians have no grounds on which to "legalise" theft.
Install bonfire.cafe
Get the full app experience
@EUCommission If everything that’s made with it is free of copyright and available as free and open source. Not just the model weights; all of it. So Antrophic, OpenAI, Google, Microsoft, etc., are forced to release all their software as free software.
In other words, if we get rid of copyright altogether for everyone and all tech and all information enters the commons. Of course, that would also take implementing universal basic income, universal healthcare, housing, etc. (i.e., a proper social safety net).
(You’ll never legislate for anything remotely approaching that. You’re just shilling for Big Tech. And no matter what anyone says, you’re going to do what’s in your neoliberal interests, like always. Like you’re doing with Chat Control.)