The EU AI Act's transparency obligations under Article 50 took effect on August 2, 2026, requiring companies to label AI-generated content designed to look authentic and to disclose when users are interacting with AI systems1,5,3.
The rules cover four categories: AI systems that interact directly with people, such as chatbots; systems that generate synthetic text, images, audio, or video; emotion-recognition and biometric-categorisation systems; and deepfakes or AI-generated text on matters of public interest. Synthetic content designed to look truthful must be visibly marked as AI-generated and contain a digital watermark8. AI-generated text published to inform the public on matters of public interest must be labelled unless it has undergone human editorial review and an identifiable person or entity bears editorial responsibility.
The rules apply to new AI systems on the EU market from August 2, while existing systems have an extra four months to comply. A grace period until December 2026 applies specifically to the machine-readable marking obligation for generative AI systems placed on the market before August 2. Content generated or manipulated before August 2 does not need to be marked retroactively.
The regulation binds not only major AI providers such as OpenAI, Google, and Meta but also individual creators who publish AI content in a professional or monetised capacity2. The European Commission draws the line at economic benefit: posting an AI video purely privately falls outside the scope, but doing so through a monetised YouTube channel or an Instagram account with paid partnerships qualifies the user as a "deployer" under the regulation. Employees are not personally liable; the employing company carries the duty.
Exceptions exist for obviously fantastical material — the Commission cites humans flying unaided as an example — and for clearly artistic, satirical, or fictional works.
Non-compliance carries fines of up to €15 million or 3% of a company's worldwide annual turnover.
Enforcement formally begins not only for the new transparency rules but also for provisions already in force, including prohibited AI practices, general-purpose AI obligations, and AI literacy requirements. New prohibitions on non-consensual deepfakes and AI-generated child sexual abuse material have been postponed to December 2, 2026, following the AI Omnibus regulation that entered into force on July 27, 2026. High-risk AI rules under Annex III have also been pushed further down the calendar.
Several major platforms have already taken steps toward compliance. Google launched SynthID in 2023 and says it has applied invisible digital watermarks to more than 100 billion images and 60,000 years' worth of audio. Meta launched policies in 2024 to identify AI-generated photo-realistic images and says it has been developing industry-wide technical standards to identify AI audio and video. TikTok says it has required creators to label realistic AI-generated content for several years and has helped label more than 3 billion pieces of content. Google and Meta have signed a voluntary code of practice intended to help companies comply with the EU's transparency rules, with more than 180 organisations having signed the code. Meta's vice-president for public policy in Europe, Markus Reinisch, said Meta intends to sign the EU code of practice.
The EU executive has created standardised black-and-white labels that anyone can use. The code was drawn up by independent experts under the supervision of the EU AI Office.
"It is a matter not only of customer protection, it's also a matter of democracy protection," said Green MEP Sergey Lagodinsky, who helped negotiate the AI Act. Wired reported concerns about "disclosure fatigue" as the volume of required labels increases4.
ANALYSIS The scope of the transparency rules — extending to individual monetised creators, not just large AI providers — broadens the compliance surface well beyond the companies that typically dominate AI-regulation discussions. The staggered enforcement timeline, with machine-readable marking and high-risk rules deferred to later dates, creates a multi-phase compliance calendar that companies operating in the EU will need to track through at least the end of 2026.