EU AI Act Takes Effect August 2: Mandatory Labels and Watermarks for AI‑Generated Content
From August 2, the EU AI Act requires clear visible labels and digital watermarks on AI‑generated or AI‑manipulated content, with heavy fines for non‑compliance.
The EU’s landmark Artificial Intelligence Act will begin to apply across member states on August 2, 2026, bringing mandatory transparency requirements for content created or modified by artificial intelligence. The law requires providers and deployers to ensure that users are informed when they are interacting with chatbots or consuming media that is AI‑generated or convincingly altered. (ai-act-service-desk.ec.europa.eu)
New labelling and watermarking obligations
From the enforcement date, certain AI outputs must carry visible labels and machine‑readable markers so that both humans and automated systems can identify generated or manipulated material.
The European Commission has published a voluntary Code of Practice to help companies implement the marking and labelling requirements in a consistent way, including an optional set of EU icons for public use. (digital-strategy.ec.europa.eu)
Platforms, publishers and generative AI providers will need to adopt a mix of visible indicators, metadata and robust watermarking solutions that survive common file transformations.
The rules target deepfakes and AI content published on matters of public interest, while also covering conversational agents whose outputs may be mistaken for human responses. (digital-strategy.ec.europa.eu)
Scope of the transparency rules and exemptions
Article 50 of the regulation sets out the transparency duties that apply to a broad range of AI systems; not all AI products face the same obligations.
A limited grace period is provided for systems placed on the market before August 2, 2026, but the marking and detection obligations will apply in key cases without delay. (digital-strategy.ec.europa.eu)
National authorities and the EU’s oversight bodies will interpret the obligations in light of the Commission’s guidance and forthcoming technical standards, which aim to strike a balance between consumer protection and operational feasibility.
The legislation draws a clear line between content that must be disclosed and routine algorithmic uses that do not materially change the nature of media consumed by the public. (digital-strategy.ec.europa.eu)
Enforcement powers and financial penalties
The AI Act creates a tiered enforcement and penalty framework, with significant fines for serious breaches of the law.
For many transparency and deployer duties, administrative fines can reach up to €15 million or 3% of a company’s total worldwide annual turnover, whichever is higher. (agentliability.eu)
More severe breaches, such as prohibited AI practices set out in the Act, carry higher caps of up to €35 million or 7% of global turnover.
Authorities will consider the nature, gravity and duration of the infringement when calculating sanctions, and non‑monetary remedies may also be imposed. (agentliability.eu)
Practical steps companies must take now
Companies operating in or offering services to EU users should inventory systems that generate or modify audio, images, video or text and map them against Article 50 transparency duties.
Technical measures will include embedding machine‑readable metadata, visible icons or captions and robust watermarking that withstands routine editing and distribution. (digital-strategy.ec.europa.eu)
Legal and compliance teams must update user‑facing terms, bot disclosures and third‑party agreements to make clear when AI is in use and to document the methods used for marking and watermarking.
Providers of general‑purpose AI models should pay particular attention to governance obligations introduced earlier in the Act’s staged roll‑out and coordinate with platform partners on detection and reporting mechanisms. (digital-strategy.ec.europa.eu)
EU oversight, timelines and future guidance
The Act’s application follows a phased timeline intended to allow for the development of standards and enforcement capacity at national and EU level.
Several secondary instruments and implementation guidelines remain under development and will be published to clarify technical specifications, including accepted watermarking formats and detection protocols. (ai-act-service-desk.ec.europa.eu)
The European Artificial Intelligence Office and national market surveillance authorities will share responsibility for monitoring compliance, while the European Commission will continue to issue guidance to reduce fragmentation and help companies meet the new obligations.
Businesses are advised to track updates and engage with industry guidance to ensure interoperability of labels and avoid enforcement risks. (ai-act-service-desk.ec.europa.eu)
The EU AI Act marks a major regulatory step aimed at curbing digital misinformation and strengthening transparency around automated content. Companies with EU market exposure should move quickly to assess their exposure, adopt resilient marking and watermarking techniques, and document compliance measures ahead of the August 2 deadline.