AIPolicyTracker
Legal requirement Transparency and disclosure European Union Partially applicable

Providers of generative AI must mark synthetic output as artificially generated in a machine-readable way

Context fileUnder EU AI Act, Article 50(2)

Source-linked Open official source

What does it require?

Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text must ensure the output is marked in a machine-readable format and detectable as artificially generated or manipulated. The technical solution must be effective, interoperable, robust and reliable as far as the state of the art allows, taking account of content type and cost. Systems that only perform assistive editing or do not substantially alter the input, and law-authorised criminal-detection uses, are outside the duty.

Practical action

Implement watermarking or content credentials on generated output and keep test evidence that the mark survives common transformations.

Who does it apply to?

Providers of systems that generate synthetic media or text, whatever the risk tier; codes of practice under Article 50(7) will detail acceptable techniques.

Applies from:

Which controls meet this duty?

Satisfies: the control, operated properly, does the work the duty asks for. Supports: it contributes but the duty needs more. Each control page lists every other duty it serves, so work done once can be counted once.

  • satisfiesTechnical measureEngineering lead · continuous
    Synthetic content labelling and provenance marking

    Serves 5 recorded duties · evidence: Content labelling and provenance standard, Watermark and provenance robustness test, Visible AI-generated content label

    Machine-readable marking with robustness evidence.

  • supportsTechnical measureQuality or testing lead · at launch and on material change
    Accuracy, robustness, fairness and security testing

    Serves 15 recorded duties · evidence: Pre-release test report, Test plan and acceptance criteria, Release test sign-off

    Tests that the mark is robust and reliable.

What evidence would a reviewer expect?

Evidence examples
EvidenceTypeNotes
Provenance marking design and robustness test reportreport
Output sample with embedded machine-readable markrecord

Framework mappings

Original editorial crosswalks. They cite clause numbers only and reproduce no standard text; confidence reflects how direct the mapping is.

See every European Union duty mapped this way →

Framework mappings
FrameworkReferenceNoteConfidence
ISO/IEC 42001:2023Annex A.8.2, A.6.2.4System documentation; verification and validation of the marking.low
NIST AI RMF 1.0MEASURE 2.7, MANAGE 4.1Provenance mechanisms as recommended in NIST AI 600-1.medium

Cite this record

AIPolicyTracker (2026). “Providers of generative AI must mark synthetic output as artificially generated in a machine-readable way (EU AI Act)”. https://aipolicytracker.org/obligations/eu-ai-act-art-50-2-synthetic-content-marking (accessed 24 September 2026). Data licensed CC BY 4.0.

Cite the official text alongside it: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence, Official Journal of the European Union, https://eur-lex.europa.eu/eli/reg/2024/1689/oj.

Similar obligations in other instruments

Informational only, not legal advice. Verify every claim against the linked official sources and consult qualified counsel before acting.