AI-generated content marking under EU AI Act Article 50

Ship AI-generated content that meets both halves of the law: machine-readable marking from the provider and visible disclosure from the deployer.

What the law requires

Article 50(2) puts machine-readable marking on the provider. Article 50(4) puts visible disclosure on the deployer: deepfakes must be labelled as artificially generated or manipulated, and AI-generated text published to inform the public on matters of public interest must be disclosed unless it passed human review with editorial responsibility.

EU AI Act Article 50(2)MACHINE READABLE MARKINGPROVIDER / externally testable
EU AI Act Article 50(4)DEEPFAKE PUBLIC INTEREST LABELDEPLOYER / externally testable
EU AI Act Article 50(5)CLARITY ACCESSIBILITY TIMINGBOTH / externally testable

What proves it

isGenAI measures the visible label, the publication workflow, the review gate and the machine-readable mark on the same output, so one scorecard shows whether a product satisfies both obligations in practice.

Official source text, public technical observations, vendor-submitted evidence and external inference are always kept apart. A missing public signal is recorded as NOT OBSERVED, a technical fact, never as a legal finding.

Observations from public scorecards

0 published observations for this topic under isgenai-methodology-2026-08-16.v1, each a dated technical observation with linked evidence. Legal applicability stays with counsel.

0Products with public scorecards
0Relevant controls verified
0Partially observed
0Not observed in tested flow

Observations publish the moment they clear the evidence gate. Output marking is verified on generated artifacts and disclosure inside the real user workflow, never from a marketing page alone.

Related public findings

Findings appear here as soon as a published scorecard records one for this topic.