EU AI Act, Article 50(4) Deepfake Disclosure Obligation

Properties

kind: law
target: voice-and-likeness
jurisdiction: EU
mechanism: Requires deployers of an AI system that generates or manipulates image, audio, or video content constituting a deepfake to disclose that the content has been artificially generated or manipulated, with exemptions for law-enforcement uses and for artistic, creative, satirical, or fictional works subject to appropriate disclosure that does not hamper enjoyment.
status: in-effect
owns: An EU-wide, deployer-facing disclosure duty that applies to any deepfake regardless of the depicted person's consent, distinct from every tracked US entity's model of a personal right of action, a consent gate, or a criminal offense.
source: EU Artificial Intelligence Act Explorer, Article 50 (Transparency Obligations for Providers and Deployers of Certain AI Systems)
notes: Article 50 applicability is governed by Article 113, which sets 2 August 2026 as the general date of application for provisions not otherwise staged. Multiple 2026 legal-alert sources (Jones Walker, Cloud Security Alliance) state the EU's 'Digital Omnibus on AI' postponed high-risk-system deadlines and gave a four-month extension (to 2 December 2026) only to the machine-readable watermarking duty in Article 50(2), leaving the Article 50(4) deepfake disclosure duty quoted here on its original 2 August 2026 schedule and therefore already in effect as of this writing; the specific claimed entry-into-force date for the Digital Omnibus itself (27 July 2026) came from an aggregated search summary rather than a document this agent read directly, so that particular date is omitted here as unconfirmed. Quote reproduced from the EU Artificial Intelligence Act Explorer site (a widely cited academic mirror of the regulation's text), not fetched directly from EUR-Lex because that fetch truncated before reaching Article 50 in this pass; treat as a close reproduction of Article 50(4) pending direct EUR-Lex confirmation.

Provenance

Source: https://artificialintelligenceact.eu/article/50/

Properties

kind

law

target

voice-and-likeness

jurisdiction

mechanism

Requires deployers of an AI system that generates or manipulates image, audio, or video content constituting a deepfake to disclose that the content has been artificially generated or manipulated, with exemptions for law-enforcement uses and for artistic, creative, satirical, or fictional works subject to appropriate disclosure that does not hamper enjoyment.

status

in-effect

owns

An EU-wide, deployer-facing disclosure duty that applies to any deepfake regardless of the depicted person's consent, distinct from every tracked US entity's model of a personal right of action, a consent gate, or a criminal offense.

source

EU Artificial Intelligence Act Explorer, Article 50 (Transparency Obligations for Providers and Deployers of Certain AI Systems)

notes

Article 50 applicability is governed by Article 113, which sets 2 August 2026 as the general date of application for provisions not otherwise staged. Multiple 2026 legal-alert sources (Jones Walker, Cloud Security Alliance) state the EU's 'Digital Omnibus on AI' postponed high-risk-system deadlines and gave a four-month extension (to 2 December 2026) only to the machine-readable watermarking duty in Article 50(2), leaving the Article 50(4) deepfake disclosure duty quoted here on its original 2 August 2026 schedule and therefore already in effect as of this writing; the specific claimed entry-into-force date for the Digital Omnibus itself (27 July 2026) came from an aggregated search summary rather than a document this agent read directly, so that particular date is omitted here as unconfirmed. Quote reproduced from the EU Artificial Intelligence Act Explorer site (a widely cited academic mirror of the regulation's text), not fetched directly from EUR-Lex because that fetch truncated before reaching Article 50 in this pass; treat as a close reproduction of Article 50(4) pending direct EUR-Lex confirmation.