Generative AI systems that were on the EU market before August 2, 2026, must mark their output in a machine-readable, detectable format by December 2, 2026, when a grace period under the EU AI Act ends. Businesses that use those tools have owed separate labeling duties for deepfakes and certain AI-generated text since August 2.
Key Takeaways
- Article 50 transparency obligations under the EU AI Act have applied since August 2, 2026, including the duty for providers to mark AI-generated content.
- Generative AI systems placed on the EU market before August 2, 2026, have until December 2, 2026, to add machine-readable marking. Newer systems have had to comply from launch.
- Businesses that publish deepfakes or certain AI-generated text on matters of public interest have owed labeling duties since August 2, with no grace period.
- About 190 organizations had signed the European Commission’s Code of Practice on Transparency of AI-Generated Content by the end of July 2026.
- Breaching Article 50 can bring fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, with the lower amount applying to SMEs and startups.
The December 2 Deadline Applies Only to Systems Already on the EU Market
The EU AI Act’s transparency rules under Article 50 began applying on August 2, 2026. The Digital Omnibus on AI, Regulation (EU) 2026/1744, which the European Parliament adopted June 16 and the Council adopted June 29, granted one narrow extension. Under Article 111(4), generative AI systems already on the EU market before August 2 have until December 2, 2026, to meet the Article 50(2) marking requirement.
Any generative AI system placed on the EU market on or after August 2 has had to comply from launch. That creates two clocks for the same rule. Newer products have been covered since summer, while products that predate the August date reach their deadline in early December. Content generated before August 2, 2026, doesn’t have to be marked retroactively. With small business AI adoption reaching 89% in 2026, many of the tools now in daily use fall in that pre-August group.
Providers Must Make AI Output Detectable by Machines
Article 50(2) applies to providers of AI systems, including general-purpose AI systems, that generate synthetic audio, images, video or text. Those providers must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. The Act requires marking solutions to be effective, interoperable, robust and reliable, as far as is technically feasible.
Separately, Article 50(1) has required providers since August 2 to tell people when they are interacting with an AI system, unless that is already obvious from the context.
Businesses Using Generative AI Have Owed Labels Since August 2
The December 2 grace period covers only providers. Businesses that use generative AI, which the Act calls deployers, have their own duties under Article 50(4), and those have applied since August 2. Deployers must disclose when images, audio or video they publish are deepfakes, and when AI-generated text is published to inform the public on matters of public interest.
A business publishing that kind of content in Europe can’t wait for its software provider’s December deadline to meet its own labeling duty. That adds a disclosure step to workflows where AI output is already a routine input, including at companies where workers spend an hour or more each week correcting AI-generated work.
Official Guidance Narrows What Must Be Marked
The Act itself exempts certain systems from the marking duty: those performing an assistive function for standard editing, those that don’t substantially alter the input data provided by the deployer or its meaning, and those authorized by law to detect, prevent, investigate or prosecute criminal offenses.
The European Commission’s Article 50 guidelines, published July 20, 2026, clarify further exclusions. Short sequences of numbers, symbols or letters fall outside the marking obligation, as does source code intended only for machine-to-machine communication and automated processing without human exposure.
A Voluntary Code Sets the Practical Benchmark
The European Commission published its final Code of Practice on Transparency of AI-Generated Content in June 2026, and the Commission and the AI Board confirmed it as an adequate way to demonstrate compliance with Article 50(2), (4) and (5). Signing is voluntary. The Code generally expects providers to use multilayered machine-readable marking and deployers to label deepfakes and qualifying AI-generated text.
By the end of July 2026, about 190 companies and organizations had signed, according to the Commission. The Code also sets a later date beyond December: signatories must implement an interoperable watermark-detection solution by February 2, 2027. For technology vendors and the managed service providers still working to show measurable returns from AI, marking and detection add a compliance layer on top of adoption costs.
Fines Reach €15 Million or 3% of Global Turnover
Article 99(4) of the AI Act sets fines for breaching Article 50 at up to €15 million or 3% of a company’s total worldwide annual turnover for the preceding financial year, whichever is higher. For SMEs and startups, Article 99(6) applies the lower of the two amounts. Prohibited AI practices carry a higher tier of up to €35 million or 7% of turnover. National market surveillance authorities and the EU’s AI Office share enforcement responsibilities.
The Rest of the AI Act Calendar Moved Further Out
The Digital Omnibus pushed back obligations for standalone high-risk AI systems, such as hiring tools and credit scoring, from August 2, 2026, to December 2, 2027. High-risk AI built into regulated products such as medical devices and machinery moved to August 2, 2028. General-purpose AI model obligations have applied since August 2, 2025, and older models placed on the market before that date have until August 2, 2027, under Article 111(3). A new prohibition on AI systems designed to generate non-consensual intimate imagery also takes effect December 2, 2026.
FAQs
What happens on December 2, 2026, under the EU AI Act?
Generative AI systems placed on the EU market before August 2, 2026, must comply with the Article 50(2) duty to mark AI-generated output in a machine-readable, detectable format.
Do businesses using AI tools have to label content under the EU AI Act?
Yes, in certain cases. Since August 2, 2026, deployers must disclose deepfakes and AI-generated text published to inform the public on matters of public interest.
What are the fines for breaching EU AI Act transparency rules?
Up to €15 million or 3% of worldwide annual turnover, whichever is higher. SMEs and startups face the lower of the two amounts.
Is the Code of Practice on AI-generated content mandatory?
No. Signing is voluntary, but the Commission and AI Board confirmed it as an adequate way to demonstrate compliance with Article 50.




