Do you have to label tattoo designs when you use AI? What the EU AI Act requires
Last reviewed: November 20, 2026. Based on the AI Act as amended by Regulation (EU) 2026/1744 and the European Commission's guidelines on Article 50 of July 20, 2026 ("Art." refers to the AI Act, "para." to the paragraph number in the guidelines). Researched and written with AI assistance, reviewed by and under the responsibility of Daniel Menius.
What has applied since August 2, 2026?
Since August 2, 2026, the transparency obligations in Article 50 of the AI Act, Regulation (EU) 2024/1689, have applied (para. 153). They don't require a label on every motif AI was involved in. Providers offer an AI system under their own name, such as OpenAI, Midjourney, or Adobe (Art. 3(3), para. 10). Deployers use an AI system under their own authority (Art. 3(4)), and that's you if you use AI for your business.
| Obligation | Who | Does it affect you as an artist? |
|---|---|---|
| Inform people that they are interacting with an AI (Art. 50(1)) | Providers | No |
| Mark outputs as AI-generated in a machine-readable format (Art. 50(2)) | Providers | No, unless you offer your own AI tool under your name |
| Disclose emotion recognition and biometric categorization (Art. 50(3)) | Deployers | No |
| Disclose deep fakes (Art. 50(4), first subparagraph) | Deployers | Yes, if you publish deep fakes |
| Disclose AI text on matters of public interest (Art. 50(4), second subparagraph) | Deployers | Rarely, for example unreviewed health texts |
So you only have to label an image if you are a deployer and it is a deep fake within the meaning of Art. 3(60).
Where does this information come from?
The basis is the AI Act as amended by Regulation (EU) 2026/1744 of July 8, 2026 (the "Digital Omnibus on AI") and the European Commission's guidelines on Article 50 of July 20, 2026 (C(2026) 5054 final). The guidelines are non-binding, and only the Court of Justice of the European Union can interpret the AI Act authoritatively (para. 5).
Anyone who signs and implements the Code of Practice on Transparency of AI-generated Content of June 10, 2026 can use it, according to the Commission, to demonstrate compliance with Art. 50(2), (4), and (5) (paras. 146 and 147). For an individual artist, it is voluntary. This article does not replace legal advice on individual cases.
This article covers the EU. The UK, the US, and other countries have their own rules, but posting publicly from outside the EU can be enough to bring you into scope (Art. 2(1)(c), para. 13).
Are you a deployer under the AI Act as an artist?
Yes, as soon as you publish AI images or AI text for your tattoo business. The only exception is use "in the course of a personal non-professional activity" (Art. 3(4), Art. 2(10)). According to the Commission, "professional" covers any activity through which someone regularly gains an economic benefit, or which is part of a business, trade, occupational, or freelance activity (para. 19). You are also a deployer if you decide on the use of AI and someone else operates the tool (para. 12).
| Situation | Deployer? | Reference |
|---|---|---|
| You post on your business account, such as flash for sale or studio reels | Yes | para. 19 |
| You create an AI image privately for your private account, with no link to the business | No | Art. 2(10), para. 19 |
| A social media helper posts AI content on your behalf and on your instructions | Yes, you remain responsible | paras. 14 and 19 |
| You hire an agency and don't decide whether or how it uses AI | No | example under para. 14 |
| Instagram or TikTok shows your image to others | The platform is not the deployer of your image | para. 16 |
| Your studio is in Switzerland and your posts are publicly visible in the EU | Yes, in the Commission's view posting publicly is enough | Art. 2(1)(c), para. 13 |
Is an AI tattoo design a deep fake?
Generally, no, because nobody takes a drawn design, a flash sheet, or a stencil for a real photo.
Art. 3(60) defines a deep fake as "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful". From this, the Commission derives four criteria, all of which have to be met (para. 113):
- Resemblance: The image appreciably resembles a real subject. It does not need to be identical.
- Existing: What is shown exists, could plausibly exist, or could plausibly have existed. Anything that defies the laws of nature is out, such as dragons or elephants driving cars.
- Category: Persons, objects (including artworks and consumer goods), places, entities (by which the Commission means non-human living beings such as animals), or events (including professional or consumer services).
- False appearance of authenticity, including as to the actual appearance of objects or the actual delivery of services.
That's why drawn designs usually fail the fourth criterion, and fantasy motifs such as demons or mythical creatures also fail the second. With realism designs, such as a photorealistic portrait or animal, photorealism alone is not decisive. What counts is whether the image could be taken for a real recording in its specific context (para. 114). If it shows a real person, data protection and personality rights apply as well, and a label doesn't make up for missing consent (paras. 127 and 129).
What counts is your foreseeable audience. If children, older people, or people with less AI experience are likely to see the image, for example on a public Instagram account, it is enough that they could be misled. Unforeseeable sharing by third parties doesn't count (para. 115).
Is an AI mockup on skin a deep fake?
Yes, if it shows a tattoo cleaner, sharper, or finer than you can achieve in skin, and could pass as real work. A tattoo on the body and a tattoo session depict a service, which makes them events within the meaning of the definition (para. 113). According to the Commission, an AI-generated product image in advertising is a deep fake if it can mislead as to the actual appearance, characteristics, or use of the product, including by making the product appear more appealing or of higher quality than in real life (examples after para. 116).
Minor edits such as color correction, lighting, noise reduction, compression, removing passersby, replacing or extending backgrounds for clearly aesthetic purposes, and re-scaling don't turn an image into a deep fake (para. 116). Nor does a real product against an AI background, as long as the ad doesn't mislead about the product itself (examples after para. 116). So your real healed photo in front of a generated background is fine, while a generated result that never existed is not.
Which AI use do you have to label as a tattoo artist?
AI images that show something real and could pass as authentic, and AI texts on matters of public interest that you haven't reviewed yourself.
| Situation | Classification | What you should do | Reference |
|---|---|---|---|
| AI gives you the idea, you redraw everything | You don't publish any AI content | Nothing | Art. 50(4) |
| Flash or design fully generated, with a clearly drawn look | Not a deep fake, no appearance of a real photo | No label needed | Art. 3(60), para. 113 |
| Generative Fill for the background of your real portfolio photo | Minor edit | No label needed | para. 116 |
| Photorealistic mockup on skin in your portfolio or ads | Likely a deep fake | Label it visibly | examples after para. 116 |
| AI mockup on a photo of a real client, published | Deep fake, plus data protection and personality rights | Get consent first and label it | paras. 127 and 129 |
| Portrait of a real person, presented as a real photo | Deep fake | Label it visibly | para. 113 |
| AI preview during a consultation, recognizable as a preview | No false appearance of authenticity | No obligation | para. 114 |
| AI caption or AI ad copy for your flash | Advertising copy, not text on a matter of public interest | No label needed | para. 131 |
| AI text on aftercare, healing, or health risks on your website | Health can be a matter of public interest | Review it yourself and take responsibility, otherwise label it | Art. 50(4) second subparagraph, paras. 131 and 133 to 138 |
| Private message or booking reply written with AI | Not published | No obligation | para. 131 |
According to the Commission, advertising copy and product descriptions are out of scope, except for claims related to health or consumer safety (para. 131). An AI text on tattoo aftercare needs no label if a human has reviewed its content and a named person holds editorial responsibility (Art. 50(4), second subparagraph). A spell check or a cursory read-through isn't enough (para. 135), and the person responsible should be easy to find, for example in your website's legal notice (para. 138).
For deep fakes in evidently artistic, creative, satirical, or fictional works, the disclosure can be designed so that it doesn't hamper the display or enjoyment of the work (Art. 50(4), para. 119). The Commission reads this narrowly: content whose nature is unclear is excluded, as is content that is recognizably purely informative or commercial, and where content mixes both, the informative character prevails (para. 122). An ad post for flash is commercial, but drawn flash isn't a deep fake in the first place.
What does an AI label have to look like?
Visible, in the image itself, and at the latest when someone first sees it. Art. 50(5) requires the information to be provided "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure" and to be accessible.
What isn't enough: anything that is easily overlooked under normal conditions, such as a notice in the terms of use, in a manual, or hidden under layers of menus (para. 142). A notice at the end of a long caption that only appears after "more" is a borderline case. The obligation applies to each output and every person, including while scrolling through social media (para. 143). In a carousel, every deep fake image therefore needs its own label.
Official icons are free. On June 10, 2026, the EU published three icons together with the Code of Practice: a basic icon for content AI was involved in, one for fully AI-generated content, and one for partially AI-modified content, as SVG and PNG. According to EU user testing, the icon is understood better with a short text next to it, such as "AI mockup". It should sit in the image and remain visible when the image is shared or downloaded. Using the icons is optional, the label itself is not.
What are the penalties, and who enforces them?
For small businesses, at most 3% of annual turnover, so up to EUR 1,800 on EUR 60,000 of turnover. In general, fines go up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher (Art. 99(4)(g)), but for small and medium-sized enterprises the lower amount applies (Art. 99(6)). Penalties have to be proportionate and take into account the economic viability of small businesses (para. 152). Member states can also provide for warnings and non-monetary measures (Art. 99(1)).
Germany: Under the KI-MIG, in force since July 29, 2026, the Federal Network Agency (Bundesnetzagentur) is the central market surveillance authority, single point of contact, and complaints office.
Austria: No market surveillance authority has been designated yet, and the RTR's AI Service Desk only gives advice. The obligations still apply, because the regulation is directly applicable.
Anyone who suspects a violation can lodge a complaint (Art. 85, para. 151), including a disappointed client or a competitor. Independently of that, a mockup that misrepresents your work may be challenged as misleading advertising (para. 129).
Old posts: You don't have to retroactively label deep fakes created before August 2, 2026. The Commission encourages it, but does not expect disproportionate effort such as going through entire archives (para. 154).
What mistakes should you avoid when labeling AI content?
Labeling everything as AI. A label on every motif dilutes the label, and honest disclosure on a real mockup gets lost.
Relying on the provider's marking. People can't perceive the invisible marking under Art. 50(2) without technical tools, so deployers cannot rely on it (para. 117).
Only thinking about Instagram. The obligation also applies to your website, Artist Page, TikTok, newsletters, and flyers.
Misreading the transitional period. The December 2, 2026 deadline in Regulation (EU) 2026/1744 only covers the technical marking under Art. 50(2) by providers whose AI systems were already on the market before August 2, 2026 (para. 153). There has never been a transitional period for deployers labeling deep fakes.
What you can try this week
Sort your last twenty posts into two piles: images that show something real (skin, faces, finished tattoos on the body, real studio scenes) and obviously drawn ones. The second pile is done. For the first, ask yourself per image: does it show a result that never existed like this, or one that looks better than what I actually tattoo? If that's more than one or two images, you have an expectations problem with your inquiries on top of the labeling problem.
How was this article made?
Researched and written with AI assistance, quotes in the official wording, every reference checked against the original regulation and guidelines. Daniel Menius reviewed the content and holds editorial responsibility, as Art. 50(4), second subparagraph, requires for AI text on matters of public interest (paras. 133 to 138).
How to cite this article
Menius, Daniel (2026): Do you have to label tattoo designs when you use AI? What the EU AI Act requires. tatme.com, last reviewed November 20, 2026. https://tatme.com/en/wissen/ki-kennzeichnung-tattoo-designs
Every section has a fixed address, for example https://tatme.com/en/wissen/ki-kennzeichnung-tattoo-designs#werkzeug-vs-autor with the overview of cases.
Changes
- November 20, 2026: Revised, sources checked against the originals, and references added for every statement.
- August 2, 2026: First published.
Sources
- Regulation (EU) 2024/1689 (AI Act), Official Journal of the EU, eur-lex.europa.eu; Article 50 in its current version at the Commission's AI Act Service Desk
- Regulation (EU) 2026/1744 of July 8, 2026 amending Regulation (EU) 2024/1689 (Digital Omnibus on AI), eur-lex.europa.eu
- European Commission: Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act, C(2026) 5054 final, July 20, 2026, ai-act-service-desk.ec.europa.eu
- European Commission: Code of Practice on Transparency of AI-generated Content, June 10, 2026, digital-strategy.ec.europa.eu
- European Commission: Commission Opinion on the assessment of the Code of Practice on Transparency of AI-generated content, July 8, 2026, digital-strategy.ec.europa.eu
- European Commission: EU Icons for labelling AI-generated content, digital-strategy.ec.europa.eu
- European Commission: Transparency obligations under Article 50 of the AI Act (FAQ), digital-strategy.ec.europa.eu
- Regulation (EU) 2024/1689, Article 99: Penalties, ai-act-service-desk.ec.europa.eu
- German Act on Market Surveillance and Innovation Promotion of Artificial Intelligence (KI-MIG), in force since July 29, 2026, gesetze-im-internet.de; Bundesnetzagentur: press release of July 29, 2026, bundesnetzagentur.de
- RTR AI Service Desk (Austria): FAQ on the AI Act (in German), retrieved November 20, 2026, rtr.at