Skip to main content

EU publishes Article 50 AI transparency guidelines

15 NOVEMBER 2026·3 MIN READ·5 SOURCES

The European Commission’s Article 50 AI transparency guidelines appeared in the Official Journal on November 12, 2026, as Communication C/2026/6005. The Commission describes them as practical guidance for authorities, AI providers and deployers.

EU publishes Article 50 AI transparency guidelines

Key takeaways · 5

  • 01

    Check whether your role is that of an AI provider, a deployer, or both: the guidelines describe distinct transparency duties for providers and deployers.

  • 02

    For covered AI-generated or manipulated outputs, assess whether people need a perceptible disclosure; machine-readable marks alone may not satisfy that duty.

  • 03

    Do not treat a superficial spell-check or grammar correction as human review or editorial control under the stated text-publication condition.

  • 04

    Research-only treatment is conditional: obligations may apply if a system is put into service or its outputs are used for other purposes.

  • 05

    If a generative system was placed on the market before August 2, 2026, note the December 2, 2026 deadline for the Article 50(2) marking and detection duty.

What the publication covers

The Commission first released the guidance on July 20, 2026, before its Official Journal publication on November 12 as Communication C/2026/6005.[1][3] The Commission presents it as practical guidance to assist competent authorities, AI providers and deployers.[2] It is nonbinding, and the Commission says authoritative interpretation of the AI Act rests with the Court of Justice of the European Union.[2] The document contains 43 pages and 155 numbered points.[3] The Commission says it will review the guidance as practical experience, market-surveillance cases and court rulings accumulate.[3]

Disclosure duties for providers and deployers

Providers must design AI systems so people are informed that they are interacting with AI.[2] They must also ensure synthetic outputs are marked in machine-readable form and detectable as AI-generated or manipulated.[2] Deployers have separate duties: the guidelines say they must inform people exposed to an AI system about its operation and disclose when content has been artificially generated or manipulated.[2] For a disclosure duty requiring a label people can perceive without specialist tools, deployers cannot rely on machine-readable marks alone.[3] The Commission says the obligations apply from August 2, 2026.[2]

Human review and research boundaries

The guidelines describe a condition for text publication: it must have undergone human review or editorial control and be subject to editorial responsibility.[2] The Commission defines human review as deliberate examination by people with relevant knowledge and professional judgment.[4] It says superficial checks, including spell-checking or grammar correction, do not qualify as human review or editorial control.[4] A research exclusion can cover systems, models and outputs developed and put into service solely for scientific research and development, but relevant transparency obligations may apply if systems enter service or outputs are used for other purposes.[1] Real-world testing is outside the separate exclusion for systems not yet on the market or in service.[1]

Scope, examples and transition date

The guidelines say open-source AI systems within Article 50 remain subject to its transparency obligations.[1] They also say non-EU deployers are covered when they foresee AI outputs being disseminated in the Union.[3] Examples include an AI-generated depiction of a celebrity influencer in a promotional context and an AI-generated product image in advertising or packaging that could mislead audiences.[3] For generative systems placed on the market before August 2, 2026, the guidance gives until December 2, 2026, for the Article 50(2) marking and detection duty.[3] The Commission says it developed the guidelines with input from Member States, the AI Board and other stakeholders through a public consultation.[5]

Teams using or distributing AI-generated material can use the guidance to identify which transparency duties may apply and where a human-visible disclosure may be needed. Providers and deployers should distinguish machine-readable marking from disclosures intended for people, while treating the document as guidance rather than a binding interpretation.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 15 November 2026

    EU publishes Article 50 AI transparency guidelines

Sources

AI fluency, one session a day, built for your work.