The €69 Compliance Tool Targeting Europe’s AI Transparency Rules

The law does not require labels on everything

The transparency obligations in Article 50 of the European Union’s AI Act took effect on 2 August 2026. They may also apply to a company outside the EU if the output of its AI is used in the EU.

An AI service that interacts with people must disclose that it is AI by the first interaction at the latest. An exception may apply when this is already obvious to an average user, but the European Commission says this exception should be interpreted narrowly.

Providers of generative AI must embed machine-readable information that identifies AI generation in images, video, audio, and text. By contrast, visible “AI-generated” labels are not required for all content. They are mainly required for deepfakes that could be mistaken for real material and certain public-interest information texts published without meaningful human review.

The cost has split into two tracks. Embed AI, a Netherlands-based provider, offers assessment and preparation services for €2,950 and €9,900. Disclo offers free scope checks, a €69 guidance bundle, and badges plus record storage for €149 per year. The depth of these services differs, so they cannot be compared directly.

The small tasks needed before legal advice

A small online store may need separate providers for legal review, translation, interface changes, and record storage just to revise one chatbot notice. That can make a few lines of text expensive enough that the business puts wording only in its terms of service or delays action.

That creates room for small services that answer separate questions: “Is our feature in scope?”, “Where should the notice go on the first screen?”, and “How do we prove the label worked last month?” Rather than replacing the full legal judgment, they handle recurring work such as screen checks, copy management, and operational records.

A Korean online store selling to European customers may not even know the name or settings of the customer-support tool it uses. Metadata can be removed during editing, resizing, format conversion, and re-uploading.

The important tool is not one that applies labels everywhere. It should distinguish between chatbots, deepfakes, public-interest information texts, and ordinary advertising images, then recommend only the actions needed. That reduces over-labelling.

These are already being sold as products

Disclo, based in the Netherlands, provides small website operators with first-chatbot notices, audio disclosure scripts, multilingual copy, and records showing that disclosures worked. It sells a €69 document bundle separately from its €149-per-year badge and record product.

Embed AI, also based in the Netherlands, helps customer-service and product teams create chatbot inventories, check disclosures by channel, assign responsible staff, and prepare materials for supervisory authorities. Its pricing ranges from a free check to a €2,950 assessment and a €9,900 preparation process.

Adobe Content Authenticity, from the United States-based Adobe, lets creators attach information about the creator and generation history to up to 50 images at once. Its public beta is free. It focuses on retaining provenance information inside files rather than solving every visible deepfake warning requirement.

Four things you could build from this

• A chatbot disclosure checker for online stores selling into Europe: A Korean online store selling cosmetics to European customers enters its address. The tool finds missing disclosures before the first chat and on screens in each language. With enforcement underway, recurring checks and screen evidence are immediately useful.

• An AI-made asset delivery checker: A production studio with 3 to 10 people sends images or video to European advertisers. It uploads a file, checks whether generation information is present, and downloads an inspection record. Whether that information survives file conversion and re-saving needs to be checked for each tool.

• A disclosure assistant for virtual-person content: An education company creating instructor or model videos that resemble real people adds visual and audio disclosures suited to the first point of exposure, then stores wording by version. For deepfakes, information inside the file alone is insufficient; people need a disclosure they can recognize.

• A repository for checking AI tools used by suppliers: A Korean company offering voice consultation features to European recruitment clients asks third-party tool providers about their disclosure method and applicable date, then stores replies, contracts, and screens in one place. Small companies may need to organize responsibility and evidence across several outside tools rather than build the technology themselves.

What to check in 30 minutes today

Choose one screen that European customers see, then record the journey from arrival through the chatbot’s first reply. Write down in one line whether an AI disclosure appears before the first reply, or whether users can learn about it only by finding the terms of service.

Why this matters where you are

Check whether your own market has rules that require AI disclosures, machine-readable provenance information, or evidence that disclosures were working. The exact obligations and enforcement structure may differ from the EU rules described here. Even so, you can identify recurring work—such as interface checks, disclosure copy, and record storage—that does not require replacing legal advice.