Blog

AI Act

EU AI Act from August 2026: what changes for AI transparency

As of 2 August 2026, the practical AI Act topic for many websites, workflows and digital employees is transparency: people should know when they interact with AI or see relevant AI-generated content.

Transparency note: this article was AI-assisted, reviewed by a human editor and is published under CODIKI's editorial responsibility.

What actually starts in August

The AI Act entered into force on 1 August 2024. The wider transparency obligations in Article 50 apply from 2 August 2026. They cover direct interaction with AI systems, machine-readable marking of AI-generated or manipulated content by providers, and disclosure duties for deployers in specific cases such as deepfakes or public-interest text without human editorial control.

This is not only about high-risk AI

The AI Act follows a risk-based model. High-risk use cases have strict requirements, but Article 50 is relevant even where no high-risk system is involved. A customer-facing chatbot, an AI assistant in a support process or a public AI-generated article can create transparency duties because people must not be misled about whether they are dealing with AI.

Do AI-generated images need a label?

The short answer is: often yes, but the legal trigger depends on the role. Providers of AI systems that generate synthetic image, audio, video or text outputs must make those outputs machine-readable and detectable as artificially generated where the obligation applies. Deployers must clearly disclose AI-generated or manipulated image, audio or video content when it constitutes a deepfake.

What counts as a deepfake

The EU explanation describes deepfakes as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear authentic or truthful to a person. A fictional role illustration is usually a different case from a fake photo of a real person, but a clear AI label avoids unnecessary ambiguity.

What CODIKI should do with example-role images

For CODIKI's illustrative role images, the prudent approach is not to limit labelling to a single blog post. AI-generated images should be labelled wherever they appear on the site. In practice, CODIKI marks the role portraits as AI-generated example images. Logos, diagrams and other non-AI assets do not need that notice.

What about AI-assisted text

For AI-generated text on matters of public interest, Article 50 contains an important exception: the disclosure obligation does not apply where the text has undergone human review or editorial control and a natural or legal person holds editorial responsibility. A practical note in an article can therefore say that the text was reviewed by a human editor. Text that is generated and published without human review would need a clear disclosure instead.

How to make this operational

The practical work is simple: inventory customer-facing AI interactions, add clear wording where a person interacts with AI, label synthetic visuals across the site, keep human editorial responsibility for published text, and document who owns each AI workflow. This is not just compliance housekeeping; it is part of trustworthy operation.

No legal advice

This article summarises official guidance available on 31 July 2026. It is not legal advice. Companies should validate concrete obligations with legal counsel, especially when AI systems affect employment, access to services, biometric data, public communication or other sensitive areas.

Sources

Next step

Turn this topic into a digital employee

CODIKI reviews recurring tasks, existing systems and useful boundaries for the first productive AI employee.

Request first assessment