🏷️ AI content is getting labels.
Just like energy ratings or safety certifications help you make informed choices, the same logic now applies to AI.
From 2 August 2026, the AI Act will require clear labelling in key cases:
🔹 Deepfakes
🔹 AI-generated or manipulated content
🔹 Interactions with chatbots and AI systems
You have the right to know whether what you see, hear or read has been made or altered by AI 🤖
@placebo @jansteen @EUCommission
Today, the code is opt-in. Come August, the legal requirements are obligated.
@[email protected] @[email protected] @[email protected] nope, useful to read further: “These transparency obligations, applicable from 2 August 2026…Even though adherence to the code is voluntary, the transparency requirements under article 50 of the AI Act are legal obligations…it provides an EU-wide recognised practical framework for signatories to demonstrate compliance with those obligations.” Opt-in now to get ready to show compliance in August.
@jansteen @stiiin @EUCommission nope, useful to read further: “These transparency obligations, applicable from 2 August 2026…Even though adherence to the code is voluntary, the transparency requirements under article 50 of the AI Act are legal obligations…it provides an EU-wide recognised practical framework for signatories to demonstrate compliance with those obligations.”
Opt-in now to get ready to show compliance in August.
@davidaugust @jansteen @EUCommission The Code, which is a guideline on how to label AI slop, is voluntary and will not become mandatory in August.
In August, the transparency obligations that were already in the AI Act become mandatory. Each member state has to choose for itself how supervision works. Given the lackluster enforcement of the GDPR, due to a lack of resources for the DPAs, I don't think there will be effective enforcement.
@Vonskinnback @davidaugust @jansteen @EUCommission In 2025, the Dutch DPA had maybe 10 employee-hours on average to spend for each reported breach and complaint (if you factor in their 36-hour work week and a bunch of paid leave; ignoring the fact that not all employees process these signals full-time or even at all, and that not all worked hours are productive). https://www.autoriteitpersoonsgegevens.nl/over-de-autoriteit-persoonsgegevens/feiten-en-cijfers
Of course, that average doesn't mean much for the long tail. I'm sure that once you're being investigated, they spend a lot of time and effort on your case. That said, if tens of thousands of signals per year result in only a few dozen investigations, that's more likely due to a bottleneck than the nature of the personal data breaches.
Credit where credit is due
@EUCommission
Now make use of it illegal in all merchandice and adverticement unless the company can prove they trained their LLM with only their own data which theh fully own the copyright for.
Also ban public LLM's (also known as Generative AI) for its proven copyright infringement of unfathomable quantity that should not be in the hands of consumers of companies. Fight the erosion of information, ownership, and the very ability for humans to think for themselves.
Great move, but somehow that will not work for malignant toxic actors that will use AI to deceive us.
An electronic signature indicating the traceable source would be a start.
Indicate the safe and approved source… blockchain could do that in a unique way.
@EUCommission A self-enforced labelling system is the same as not doing anything: the ones who were transparent will use the tag, but the ones who weren't will move on without disclosing AI use.
Please stop beating around the bush and forbid GenAI that's been trained with stolen data and regulate datacenters already.
the CE label really isn't the good example you think it is
also energy labels? I'm all for those energy classes and normed labels… but then there was that A++++-Fuckup?
…and that was solved by making the previous A into a C and the previous A++ into A? And the label looked the same? Because making it distinct would be too obvious?
All I'm saying is: your idea might be okay, but the examples are stupid
No, but the label would make it easy to call out the deepfake as being fake.
Which, honestly, is probably better. Banning things rarely works and tends to hurt innocent people as a side effect (see also: social media age verification). Labels, on the other hand, hurt only those who are trying to deceive.
As for enforcement, I dunno. Malicious actors could post a malicious deepfake and crop the label off. But if they do, then they're broadcasting that they've done something illegal.
@EUCommission “voluntary”….
How stupid do you think we are?
@voxel @EUCommission it says “providers who sign”. It’s a voluntary certification program which doesn’t even cover individuals/independent parties using gen AI on their own hardware.
Read better.
@CynAq @EUCommission Not sure if you're trolling, but if you do what I suggested you would see that: «(...) to help providers and deployers of generative artificial intelligence (AI) systems meet the AI Act transparency obligations that will apply from 2 August 2026.
From that date, the AI Act will require clear labelling in key cases. Deepfakes and AI-generated or AI-manipulated text published on matters of public interest must be clearly labelled.» (emphasis added).
@voxel @EUCommission the same question back at you. It clearly says “…who sign”. Parties don’t sign anything are not obligated to do anything because they have nothing to lose without the compliance certificate they clearly rejected to begin with.
Second, I think you understand that this has nothing to do with protecting individuals from deep fakes, plagiarism etc from other individuals not using a commercially available service but running models on their own private hardware, and only addresses commercial services and publications that have some public facing official interface.
@CynAq @EUCommission I will no longer engage in this conversation as you seemingly do not seek to engage in a meaningful discussion and reject every clarification.
For anyone else, read the legal text of the AI Act.
In particular: Art. 3 (definitions) and Art. 50 (transparency obligations)
Dustin
European Commission
Jan Steen
Holger Krupp
Grey Pilgrim
Aleksei 🇪🇪
David August ❌👑
Stijn van Drongelen
Baron Vonskinnback
Michael Westergaard
Michiel
Felipe 
Jan-Willem Markus
O.Voão
satmd
Samir
canleaf08✅ 加拿大葉子 🏳️🌈⚧️🇪🇺🍁
ekari
Mark Gjøl
rzeta0
Niels J
Hiisikoloart
xs4me2
D4v
EF
Spookybot
Laurielle
Seiðr
Cegorach
Sascha
Voxel
senda 
dieTasse
ARGVMI~1.PIF
Carola 
CynAq🤘