top of page

Deepfakes And The Compliance Vacuum: Labelling AI-Generated Content Under Article 50 Of The EU AI Act

Oct 10
2 min read



Dr. Deepika Rani, Dr. Jay Kumar Pandey and Dr. Rajendra Kumar


ABSTRACT


Article 50 of the EU Artificial Intelligence Act requires providers to mark synthetic content “in a machine-readable format” so that it is “detectable as artificially generated or manipulated”, and requires deployers to disclose deepfakes. It applies from 2 August 2026, yet it leaves its operative terms to “the generally acknowledged state of the art, as may be reflected in relevant technical standards”. This article asks when AI-media duties actually bind, and who fixes what they require. It codes forty-seven obligations drawn from fifteen instruments in five jurisdictions at two snapshot dates, 1 January 2026 and 1 January 2028, with every commencement date verified against a primary text. Thirty obligations (63.8 per cent; 95% CI 49.5–76.0) were operative on 1 January 2026, and sixteen were enacted but not yet in application. Independently of the calendar, between five and fifteen of the thirty operative duties carry a residual constraint that commencement does not cure, depending on whether an author-coded “undefined term” flag, which an independent automated coder did not reproduce, is admitted. Doctrinally, the article shows that Article 50 combines delegation with no presumption of conformity: harmonised standards and common specifications under Articles 40 and 41 do not reach it, the legislature has stated that the codes of practice on which it relies “do not grant a presumption of conformity”, and the Commission has said that adherence to the Code “does not constitute conclusive evidence of compliance”, while telling supervisors to focus on whether signatories adhere to it. On the reasoning of James Elliott, a private specification cannot therefore acquire the status of EU law that a harmonised standard has. A paired contrast with China, where a mandatory labelling standard accompanied the duty, shows that this architecture is a design choice. The article closes with remedies available to the Commission and to regulated parties.


Keywords: AI Act; Article 50; deepfakes; labelling of AI-generated content; labeling; content provenance; C2PA; harmonised standards; presumption of conformity; regulatory delay.



Indian Journal of Law and Legal Research

​

Abbreviation: IJLLR

​

ISSN: 2582-8878

​

Website: www.ijllr.com

​

Accessibility: Open Access

​

License: Creative Commons 4.0

​

Submit Manuscript: Click here

​

Licensing: ​

 

All research articles published in The Indian Journal of Law and Legal Research are fully open access. i.e. immediately freely available to read, download and share. Articles are published under the terms of a Creative Commons license which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

 

Disclaimer:

​

The opinions expressed in this publication are those of the authors. They do not purport to reflect the opinions or views of the IJLLR or its members. The designations employed in this publication and the presentation of material therein do not imply the expression of any opinion whatsoever on the part of the IJLLR.

bottom of page