A transitional period under the EU’s artificial intelligence regulation ends on December 2: from that date companies that provided AI-generated content before August 2, 2026 must also comply with transparency rules. EY warns that the change may affect Hungarian software developers, businesses embedding AI features into their products, and firms offering or using AI solutions under their own name.
What the EU rules require
One key objective of the regulation is to make content created or materially modified by AI identifiable — this covers text, images, audio and video. Identification can be achieved through provenance information embedded in files, metadata, digital watermarks or other machine-readable markers. The EU does not mandate a single technology; the requirement is that the chosen solution must reliably support recognition of AI-origin content in practice.
Who is covered by the rules?
The regulation has applied to systems placed on the market since August 2, 2026, while tools that were already available in the EU before that date were granted a four-month transitional period. That period expires on December 2. In Hungary this means not only large international AI platforms need to address the change: any domestic company could be affected if it uses or sells software under its own name that generates content with AI. The obligation also extends to firms that incorporate or rebrand third-party developer technology as their own.
A company’s responsibilities depend on how it uses the AI solution: in some situations the same organization may act as a user, in others as a provider.
Technical and organisational obligations
Compliance goes beyond adding a technical marker. Companies should check whether provenance information remains attached after downloads, compression, transformation or uploads to other systems. Resizing an image, reformatting a video or uploading a file to a platform can remove embedded data, so measures are needed to ensure the label persists.
EY notes that preparation is not solely an IT task: legal, procurement, marketing, communications and content teams may need to cooperate, especially where a firm adapts or republishes content from external AI providers. Zala Mihály, partner responsible for EY’s IT and Technology Advisory services, advised companies to review before the December deadline which relevant technologies they offer, which they use only internally, and what happens to the content produced by them.
Visible labels versus machine-readable markers
The December deadline primarily concerns the machine detectability of AI content; it does not necessarily require a prominent visible warning on every AI-generated item. However, in some cases — for example when publishing deepfake-like, convincingly artificial or altered image, audio or video — the publisher must clearly inform users.
For AI-generated text, visible labeling is not automatic: factors include the purpose of publication, the role of AI in producing the text, whether human review took place, and whether a person or organisation assumes editorial responsibility for the content.
Why this matters for companies
The new requirements are not intended to ban AI use but to help people, businesses and online systems more easily recognise artificially produced or modified content. EY emphasises that adequate preparation can also protect digital trust and corporate reputation, so companies should audit and adjust their processes and technologies in time.
