The short answer
As of August 2, 2026, Article 50 of the EU AI Act is enforceable. For graphic designers, this means any AI-generated or significantly manipulated image, video, or audio that could be mistaken for real must be marked as artificial in a machine-readable format. This applies to work seen by an EU audience, regardless of where you are located. [/ANSWER–BOX] The European Union’s AI Act is now law, and its transparency rules—specifically Article 50—became legally binding on August 2, 2026. This isn’t a future problem; it’s a present-day compliance requirement for working professionals. If you use generative AI tools to create images for clients and that work is seen by people in the EU, this law applies to you. This isn’t a ban on AI tools. Instead, Article 50 is a “truth-in-labeling” rule. Its goal is to ensure people know when they are interacting with AI or viewing AI-generated content. For professional designers, this introduces a new step in our workflow: ensuring provenance. This guide breaks down what the law requires, how to comply using current tools, and what it means for your client deliverables. ZEKAI reviews tools independently to help you navigate changes like this one and build a future-proof workflow, which you can explore further in our hub for AI in graphic design.
What Article 50 Actually Requires of Designers
Article 50 creates distinct obligations for “providers” of AI systems (the companies making the tools) and “deployers” (people like us who use the tools in a professional capacity).
As a designer using AI tools for client work, you are a “deployer.” Your core obligations under Article 50(4) kick in when you generate or manipulate content that falls into two main categories:
- Deepfakes: Any image, audio, or video that resembles real people, objects, places, or events in a way that could falsely appear authentic. The EU’s definition is broad and includes realistic product shots, AI-generated backgrounds, and synthetic characters.
- Public Interest Text: AI-generated text published on matters of public interest. This is less common for visual design work but could apply to ad copy or social media campaigns addressing health, environmental, or political topics.
For this content, you must clearly disclose that it has been artificially generated or manipulated. However, a simple “Made with AI” caption isn’t enough. The law’s implementing guidelines, the Code of Practice, favor a multi-layered approach.
The “providers” of AI tools, like Adobe and OpenAI, have their own requirement under Article 50(2): they must ensure their outputs are marked in a machine-readable format so they are detectable as AI-generated. The technical solution must be “effective, interoperable, robust, and reliable.” This is where the C2PA standard comes in.
3% Source: holisticai.com
The maximum fine for violating the AI Act’s transparency obligations, including Article 50, is up to €15 million or 3% of worldwide annual turnover, whichever is higher.
The C2PA Standard: Your Compliance Toolkit
The EU AI Act does not explicitly name any specific technology for compliance. However, the European Commission’s Code of Practice—the official guide for implementing the law—points to the C2PA (Coalition for Content Provenance and Authenticity) standard as the preferred method.
C2PA is an open technical standard that attaches a tamper-evident manifest of a file’s origin and history directly to the asset itself. This manifest is called a Content Credential. It functions like a digital nutrition label, showing:
- Who (or what AI tool) created the file.
- When it was created.
- Any subsequent edits or modifications.
C2PA
The de facto technical standard for Article 50 compliance, supported by major toolmakers.
The de facto technical standard for Article 50 compliance, supported by major toolmakers.
C2PA is not a product you buy; it’s a standard that toolmakers implement. Major companies like Adobe, Microsoft, and OpenAI are C2PA founders and have integrated Content Credentials into their flagship generative tools. When you generate an image with a compliant tool, the Content Credential is automatically embedded.
- Price from
- Free open standard
- Free tier
- N/A
How Major AI Tools Handle Article 50 Compliance
Adoption of C2PA is uneven across the most popular AI image generators. For a professional workflow that prioritizes compliance, choosing a tool with built-in Content Credentials is the most direct path. As of September 2026, here’s where the major platforms stand.
| Tool | C2PA Support | How it Works | ZEKAI Review |
|---|---|---|---|
| Adobe Firefly | Yes | Automatically embeds Content Credentials in all generated images, from the web app to Photoshop’s Generative Fill. | [Coming Soon] |
| DALL-E 3 (via ChatGPT/API) | Yes | OpenAI began adding C2PA Content Credentials to DALL-E 3 outputs in 2023. | [Coming Soon] |
| Midjourney | No | As of mid-2026, Midjourney does not embed C2PA credentials or any other standard watermark, despite being a member of the Content Authenticity Initiative. | /ai-tools/midjourney/ |
| Google Gemini / Imagen | Yes | Google uses a combination of C2PA credentials and its own invisible SynthID watermark. | [Coming Soon] |
Swipe the table sideways →
The Midjourney Gap
Midjourney
Top-tier image quality but a major compliance risk due to its lack of C2PA support.
Top-tier image quality but a major compliance risk due to its lack of C2PA support.
For designers concerned with Article 50, the lack of provenance data in images from Midjourney is a significant problem. An image downloaded from Midjourney carries no machine-readable information about its AI origin. This forces deployers (you) to handle the entire disclosure burden manually and makes it difficult to prove compliance if audited. Note that the EU’s 2026 AI Omnibus agreement grants generative AI systems already on the market before August 2, 2026 a grace period until December 2, 2026 to meet the provider-side machine-readable marking duty under Article 50(2) specifically — but this does not touch your own deployer obligations under Article 50(4) to disclose deepfakes and AI-generated content, which still apply from August 2, 2026. Given the regulatory environment, building a professional workflow around a tool with no built-in provenance remains a risky strategy.
As of September 2026, Midjourney has no free trial. Paid plans start at the Basic Plan for $10/month (or $8/month annually) and go up to the Mega Plan at $120/month (or $96/month annually). Only the Pro ($60/mo) and Mega ($120/mo) plans offer “Stealth Mode” to keep generations private.
- Price from
- $10-$120/mo
- Free tier
- No free tier (as of Sep 2026)
What About Other Tools?
Tools that are not image generators, like the AI color palette tool Khroma, are not directly impacted by Article 50’s image-marking rules. Khroma helps you generate color palettes based on your personal taste and is completely free to use. Since it doesn’t generate synthetic media that could be mistaken for “real,” its output doesn’t require C2PA credentials.
A Compliant Workflow for AI-Assisted Design
Integrating Article 50 compliance into your workflow requires a “provenance-aware” mindset.
- Generate with a Compliant Tool: Start your process in a tool that automatically embeds C2PA Content Credentials, like Adobe Firefly or DALL-E 3. This fulfills the provider’s core obligation and gives you a verifiable starting point.
- Edit in a Compliant Editor: When you open the AI-generated asset in a C2PA-aware program like Adobe Photoshop, the software will read the existing Content Credential. When you save your edits, Photoshop appends your changes to the credential, creating a secure log of the entire creative process.
- Disclose at Deployment: When you deliver the final asset to a client or publish it online, you still have an obligation as the deployer to disclose its AI origins, especially if it qualifies as a “deepfake” under the EU’s broad definition. This can be a clear label in the context where the image is displayed. The embedded C2PA credential serves as the underlying technical proof to back up your visible disclosure.
- Educate Your Clients: Inform your clients about the AI Act and explain that providing provenance data is now a legal requirement, not an optional extra. Frame it as a form of professional due diligence that protects both you and them.
As part of our compliance with the EU AI Act (Article 50), which is now in effect, all AI-generated assets we deliver include embedded Content Credentials. This is a secure digital record verifying the asset's origin and any subsequent edits. It ensures we both meet our legal transparency obligations for any work displayed to an EU audience.
US vs. EU: A Quick Comparison on AI & Copyright
It’s crucial not to confuse the EU’s provenance rules with US copyright law. The two address different issues.
- EU AI Act (Article 50): Focuses on transparency and provenance. It requires you to disclose that an image is AI-generated. It doesn’t rule on who owns it.
- US Copyright Office: Focuses on authorship and ownership. As of 2026, the USCO holds that works generated purely by AI with no significant human creative input are not protectable by copyright and fall into the public domain. Merely writing a prompt is not considered sufficient human authorship.
This means an image generated with Firefly can be both fully compliant with the EU AI Act (because it has C2PA credentials) and simultaneously be in the public domain under US law (if your creative input was minimal). For designers, this underscores the importance of adding substantial human modification and creativity to any AI-generated base asset if you intend to claim copyright in the final work.
As a working designer, you must now consider both provenance (for EU compliance) and authorship (for US copyright). This new legal layer is a fundamental shift in our professional responsibility, and staying informed is the best way to protect your work and your clients. Explore our full AI for graphic design hub to keep up with the tools and workflows that are shaping the future of the industry.
Where to go next
Three routes, picked for what you just read.
What is EU AI Act Article 50?
Article 50 is the transparency section of the EU AI Act. It requires that providers and deployers of AI systems inform users when they are interacting with AI. For graphic designers, it specifically mandates that synthetic content like images, audio, or video be marked in a machine-readable format so it can be identified as AI-generated.
When does Article 50 of the EU AI Act take effect?
The transparency obligations under Article 50 became legally binding and enforceable on August 2, 2026. This was 24 months after the main EU AI Act entered into force. Any AI-generated content created for an EU audience after this date must comply with the marking and disclosure rules. One narrow exception: the 2026 AI Omnibus agreement gives generative AI systems already on the market before that date until December 2, 2026 to implement the machine-readable marking duty under Article 50(2) specifically — deployer disclosure duties under Article 50(4) are unaffected and still apply from August 2, 2026.
Does the EU AI Act require C2PA?
No, the law itself is technology-neutral and does not explicitly name C2PA. However, it requires a marking solution that is robust, reliable, and interoperable. The European Commission’s official Code of Practice, which guides implementation, cites C2PA as the leading technical standard that meets these requirements.
What is the penalty for not complying with Article 50?
Non-compliance with the transparency obligations in Article 50 can result in administrative fines of up to €15 million or 3% of a company’s total worldwide annual turnover from the preceding financial year, whichever is higher. For providing incorrect or misleading information, fines can reach €7.5 million or 1% of turnover.
Does Article 50 apply to companies outside the EU?
Yes. The EU AI Act applies to any provider or deployer whose AI-generated content is intended for use or is viewed within the EU. This means a US-based graphic designer creating an ad campaign for a client that runs in Europe must comply with Article 50’s transparency rules.
Is AI-generated art copyrighted in the EU?
Article 50 does not directly address copyright. It is a transparency law. Copyright of AI-generated works in the EU is handled by existing copyright directives and national laws, which, like in the US, generally require human authorship for protection. The focus of Article 50 is on disclosure, not ownership.
What is the difference between a “provider” and a “deployer”?
A “provider” is the entity that develops an AI system and places it on the market (e.g., Adobe, OpenAI). A “deployer” is an entity that uses an AI system under its own authority in a professional capacity (e.g., a graphic design agency using Firefly for a client project). Both have obligations under Article 50.
Where to go next
Three routes, picked for what you just read.
Sources (45)
- https://www.whitecase.com/insight-our-thinking/long-awaited-eu-ai-act-becomes-law-after-publication-eus-official-journal
- https://www.holisticai.com/blog/penalties-of-the-eu-ai-act
- https://wpseoai.com/blog/c2pa-explained-the-watermarking-standard-behind-the-eu-ai-act/
- https://lumethic.com/2026/07/13/ai-image-watermarks-in-2026-dall-e-midjourney-firefly-grok/
- https://www.adlawaccess.com/2026/04/articles/eu-ai-act-and-c2pa-what-article-50-requires-for-ai-content/
- https://www.traverssmith.com/insights/knowledge/2026/is-it-a-bot-eu-ai-act-transparency-rules-take-effect-2-august-2026/
- https://www.openasset.com/blog/a-guide-to-image-copyright-and-fair-use-images-in-2026
- https://www.adlawaccess.com/2026/07/articles/eu-ai-act-guidance-expands-ai-disclosure-rules-for-advertisers-and-pr-teams/
- https://shibolet.com/law-firm-in-israel-shibolet/updates/eu-ai-act-article-50/
- https://www.archibal.com/blog/eu-ai-act-article-50-august-2-2026-creative-work
- https://www.orrick.com/en/Insights/2026/08/EU-AI-Act-Transparency-Obligations-for-AI-Generated-Content-Article-50
- https://www.elevenlabs.io/blog/the-eu-ai-act-what-generative-ai-companies-need-to-know-in-2026/
- https://www.orrick.com/en/Insights/2024/09/The-EU-AI-Act-Oversight-and-Enforcement
- https://www.normcyber.com/blog/penalties-for-non-compliance-and-governance-under-the-eu-ai-act/
- https://www2.deloitte.com/us/en/pages/consulting/articles/eu-ai-act.html
- https://garrettham.com/can-ai-generated-works-be-copyrighted-2026-legal-guide/
- https://copyrightandcreativity.org/2026/04/30/can-you-copyright-ai-generated-content-the-complete-2026-guide/
- https://uxplanet.org/10-ai-tools-that-can-10x-your-ui-ux-workflow-free-paid-38c2a967fa7e
- https://www.fotoware.com/blog/what-is-c2pa
- https://www.guideflow.com/blog/ai-design-tools
- https://www.aitoolcurator.com/tool/khroma/
- https://www.pebblebed.com/blog/c2pa-content-provenance-regulations-2026
- https://fredericksburg.com/pr/business/three-copyright-rulings-and-an-eu-deadline-have-rewritten-the-rules-for-ai-images/article_33458c3a-156c-11ef-9316-373d3170e28f.html
- https://medium.com/@joseph.desmond.cruel/how-to-understand-article-50-of-the-eu-ai-act-production-side-963c64c73b06
- https://www.artificialintelligenceact.eu/article/50/
- https://www.gamsgo.com/blog/how-much-does-midjourney-cost
- https://www.networksolutions.com/blog/2026/03/the-best-ai-tools-for-graphic-design/
- https://www.dudkowiak-putyra.com/en/ai-act-article-50-ai-content-labeling-and-code-of-practice-from-august-2-2026/
- https://www.penderlaw.com/news-and-events/who-owns-the-rights-to-ai-generated-images-a-straightforward-guide-to-copyright-law
- https://www.techsmith.com/blog/what-counts-as-ai-generated-content-under-the-eu-ai-act/
- https://www.designers-choice.com/reviews/khroma
- https://blog.pageproof.com/what-are-content-credentials/
- https://whytryai.com/midjourney-pricing/
- https://wedetect.it/detection-guide
- https://www.pebblous.com/blog/eu-ai-act-article-50-aug-2-ai-content-labeling-rules-c2pa
- https://www.contentauthenticity.org/blog/which-ai-image-generators-support-c2pa-midjourney-dall-e-3-sora-more
- https://gradually.ai/midjourney-statistics/
- https://www.evolink.ai/midjourney-pricing/
- https://blankai.app/blog/midjourney-metadata-what-is-embedded-and-how-to-remove-it
- https://helpx.adobe.com/firefly/faq.html
- https://clearwatermark.com/blog/ai-watermark-comparison
- https://www.cabina.ai/blog/how-much-does-midjourney-cost
- https://business.adobe.com/blog/perspectives/content-authenticity-arrives-for-enterprises
- https://blog.adobe.com/en/publish/2023/04/03/cai-1000-members-content-credentials-adobe-firefly
- https://helpx.adobe.com/creative-cloud/help/content-credentials.html
See Zekai first in Google
The weekly AI briefing for your profession
One weekly email: the AI changes that actually affect your profession — tools, deals, and what to do about them.


