EU AI Act enforces new transparency rules for AI‑generated content

EU AI Act transparency AI labeling watermarks AI literacy

From 2 August 2026 the European Union’s AI Act entered a new phase that obliges companies to clearly label AI‑generated or AI‑manipulated images, video, audio and text. The labeling must be done through visible watermarks, metadata or other digital signatures so that users can recognise synthetic material. The rule also requires that users be informed when they are interacting with chat‑bots, emotion‑recognition or biometric‑categorisation systems, unless the interaction is already evident.

Exemptions apply to routine post‑production edits such as colour correction, brightness adjustment or minor retouching, as well as to artistic, fictional or satirical content. The Act also mandates AI‑literacy training for employees who use AI tools and introduces fines and sanctions for non‑compliance. Small and medium‑size enterprises receive relief measures, including vendor‑responsibility provisions, free regulatory sandboxes and simplified documentation templates.

In parallel, a coalition of major technology firms (Adobe, Microsoft, Google, Intel, Sony, BBC and others) is promoting an open “Content Credentials” protocol that embeds cryptographically signed provenance data in AI‑created media, helping platforms verify authenticity and integrity.

Entities

AI Act · Andrea Moletto · C2PA (Coalition for Content Provenance and Authenticity) · Calero · Companies · European Commission · European Commission’s AI Office · European Union · Frank Fischer · Google · Juan Pablo Figueroa · OpenAI

Claims

What the coverage asserts, and how many sources carry each claim.

[● 4 SOURCES] The AI Act classifies AI systems into risk levels: minimal, limited, high and unacceptable. www.futuro.cl · www.campaniadih.it · www.e-commerce-magazin.de · www.deia.eus
[○ 1 SOURCE] The European Commission’s AI Office now has powers to supervise generative‑AI models and impose fines for non‑compliance. www.deia.eus
[○ 1 SOURCE] From 2 December 2024 AI systems must incorporate watermarks for manipulated text, images, video and audio that could mislead the public. www.deia.eus
[● 2 SOURCES] Companies must ensure AI literacy training for employees who use AI tools. www.narrador.es · www.campaniadih.it
[○ 1 SOURCE] Chatbots must inform users they are interacting with AI unless the interaction is already evident. www.campaniadih.it
[○ 1 SOURCE] The AI Act prohibits high‑risk practices such as subliminal manipulation, targeting vulnerable groups and social‑scoring. www.narrador.es
[○ 1 SOURCE] The first part of the EU AI Act entered into force on 2 February 2025. www.narrador.es
[DISPUTED] From 2 August 2024 a new phase of the AI Act requires AI‑generated content to be clearly labelled or watermarked. www.campaniadih.it · www.deia.eus
[● 3 SOURCES] Non‑compliance with the AI Act’s transparency obligations can result in fines and sanctions. www.campaniadih.it · www.e-commerce-magazin.de · www.deia.eus
[● 3 SOURCES] Standard post‑production edits, artistic, fictional or satirical content are exempt from the AI‑labeling requirement. www.campaniadih.it · www.e-commerce-magazin.de · www.deia.eus
[● 2 SOURCES] A coalition of technology companies (C2PA) is developing an open “Content Credentials” protocol that embeds cryptographically signed provenance data in AI‑generated media. diariodelcauca.com.co · www.diariodelsur.com.co
[● 3 SOURCES] The AI Act mandates AI‑literacy training for employees who use AI tools. www.narrador.es · www.elperiodico.digital · www.campaniadih.it