EU AI Act mandates transparency for AI-generated content

EU AI Act transparency obligations digital watermarking deepfake disclosure AI content labeling

New transparency obligations under Article 50 of the EU AI Act took effect on August 2, 2026. These regulations mandate that providers of AI systems clearly label synthetic content, including images, audio, and video, to help users distinguish between human-made and machine-generated media. Specifically, AI-generated content that resembles real people, places, or events must be disclosed, and audio deepfakes require an audible disclaimer at the start.

To facilitate compliance, the European Commission has developed a set of official icons for labeling. While the use of these specific icons is optional, the underlying requirement for disclosure remains mandatory. Companies must also ensure that AI-generated text concerning matters of public interest that lacks human editorial control is clearly identified. Failure to comply with these transparency rules can result in significant financial penalties, with some reports citing fines of up to 7% of a company's global annual turnover.

In response to these mandates, companies like Anthropic have begun implementing technical solutions, such as invisible digital watermarking in text and the use of the C2PA standard for files. These measures aim to make content machine-readable and traceable without necessarily altering the user experience or the readability of the content.

Entities

Andrea Nahles · Anthropic · Banco Santander · Bremen · Bundesagentur für Arbeit · ChatGPT · Claude · European Commission · European Union · Google · Jens Blüggel · Landessozialgericht Nordrhein-Westfalen

Claims

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

[○ 1 SOURCE] Article 4 of the AI Act requires deployers of AI systems to ensure sufficient AI literacy among staff.
[● 4 SOURCES] Anthropic has implemented invisible watermarks in all new Claude model texts as of August 2, 2026. www.finanznachrichten.de · www.wp.de · www.harzkurier.de
[● 12 SOURCES] The EU AI Act entered a new phase of implementation on August 2, 2026. cyberdefence24.pl · www.eldia.es · www.diarioestrategia.cl · larevuetech.fr · www.agf.nl · +6 more
[● 5 SOURCES] Banco Santander aims to generate over 1 billion euros in business value from AI between 2026 and 2028. www.eldia.es · www.lacronicabadajoz.com · www.informacion.es · www.elperiodicoextremadura.com · www.levante-emv.com
[● 8 SOURCES] Article 50 of the AI Act mandates transparency obligations for certain AI systems. cyberdefence24.pl · www.eldia.es · www.diarioestrategia.cl · larevuetech.fr · www.agf.nl · +2 more
[● 3 SOURCES] Non-compliance with the AI Act can result in fines up to 35 million euros or 7% of a company's global turnover. www.diarioestrategia.cl · www.johner-institut.de · borncity.com
[● 2 SOURCES] Users must be informed when they are interacting with an AI system. cyberdefence24.pl · borncity.com
[● 3 SOURCES] The regulation requires AI-generated content to be machine-readable. borncity.com · larevuetech.fr · tabletzona.es
[● 3 SOURCES] AI-generated text regarding public interest that lacks human review or editorial control must be disclosed. www.europesays.com · eveningreport.nz · freddyolsson.se
[● 2 SOURCES] Audio deepfakes require a short audible disclaimer at the beginning of the content. www.europesays.com · eveningreport.nz
[● 4 SOURCES] The EU has developed a set of icons for labeling authentic-looking AI content. www.europesays.com · eveningreport.nz · freddyolsson.se · larevuetech.fr
[● 10 SOURCES] Article 50 of the EU AI Act took effect on August 2. www.europesays.com · eveningreport.nz · freddyolsson.se · tabletzona.es · cyberdefence24.pl · +5 more