Europe’s emerging AI disclosure system is moving beyond the idea of a single visible badge. Task forces supporting the European Union’s Code of Practice on Transparency of AI-Generated Content are now working on interoperable detection, provider-neutral methods, sector-specific practices, and an interactive second layer of information linked to the EU’s AI icon or an equivalent label.
For artists, that second layer may be the most consequential part. A mark can tell an audience that artificial intelligence was involved. It cannot, by itself, explain what the system did, which elements were generated or altered, whether a person’s likeness was used with consent, or where human creative decisions remained decisive.
From a symbol to an information system
The European Commission says more than ninety signatories established two implementation task forces on September 30. One focuses on marking and detection. Its current work includes interoperability, standardized interfaces, public benchmarks, red-teaming, and methods that do not depend on a single AI provider.
The second task force focuses on labels visible or audible to people. Its objectives include developing the EU icon, defining common ways to disclose the nature of AI involvement, and creating practices suited to different sectors. A working group scheduled for January 2027 will decide what information belongs in the icon’s interactive second layer and how it can connect to machine-readable markings. Another will develop an audio disclosure, or earcon, for audio-only content and accessible use.
This architecture recognizes that transparency has at least two audiences. Machines need robust signals that can survive distribution and be detected across platforms. People need an explanation they can perceive and understand without a specialized tool.
What the law does—and does not—say about creative work
Article 50 of the EU AI Act has applied since August 2, 2026. Providers of generative systems must generally design their systems so synthetic image, audio, video, and text outputs carry effective machine-readable marks, subject to defined exceptions. Deployers have separate duties for deepfakes and certain public-interest text.
The Commission’s guidance expressly recognizes artistic, creative, satirical, and fictional contexts. When an AI-generated or manipulated work qualifies as a deepfake, disclosure may be made in an appropriate way that does not interfere with the display or enjoyment of the work. Standard assistive editing can also fall outside the provider marking obligation.
That means the rules should not be reduced to the claim that every artist must place the same icon on every AI-assisted work. Obligations depend on the role of the actor, the kind of system and output, the context, and whether the content meets the legal definition at issue.
Artists can prepare a useful record now
Even where a particular label is not legally required, artists can make future disclosure easier by keeping a compact creation record. It can note the tool and version, the parts of the work affected, the source material supplied, major human transformations, collaborators, permissions for likeness or voice, and any machine-readable credentials attached at export.
The record should separate assistance from substitution. Noise reduction, translation, color correction, generative fill, synthetic performance, and a fully generated scene do not carry the same creative or ethical meaning. If the public-facing label eventually opens a second layer, that distinction is exactly the information an audience may need.
Artists should also test whether provenance data survives common edits, screenshots, compression, and platform uploads. A disclosure strategy that disappears when a gallery resizes an image or a collector saves a copy is not yet durable. Visible language and machine-readable credentials work best as complementary layers.
The EU process is still developing, and its sectoral practices may change. But its direction is already clear: the future of AI transparency is unlikely to be one universal symbol with one universal meaning. It will be a system that connects a visible signal to a more specific account of how the work was made.
ATIM approaches artificial intelligence with curiosity, independence, and balance. We explore its extraordinary potential to empower artists while examining the questions it raises around authorship, copyright, consent, creative identity, transparency, and human expression.
We are neither advocates for AI nor opponents of it. Our focus is how the technology is used—and what that means for artists.
Above all, our commitment is to artists. We look at every development through the lens of what can support, protect, and empower the creative community—identifying opportunities, examining potential risks, and providing artists with the information they need to make informed decisions about their work, their rights, and their future.