AI Notice

Draft v0.1, 2026-10-16

1. You are using an AI system

Enigma LABZ runs language models on your computer. When you chat in the App, you are interacting with an artificial intelligence system, not a person. We tell you this under Article 50(1) of Regulation (EU) 2024/1689 (the AI Act).

2. Outputs can be wrong

Language models generate text statistically. They can state false facts, invent sources or give out-of-date information. Check anything that matters, and do not treat outputs as legal, medical or financial advice.

3. Models run on your device

The Model runs on your own hardware. Your conversations are not sent to us or to the Model’s creator. Our acceleration changes how the Model is executed on your hardware; we do not alter the Model’s weights to change its behaviour.

4. How the app tells you

The first time you interact in a chat window, the App shows a clear message that you are talking to an AI system. In line with Article 50(5), it is clear, distinguishable and accessible: screen readers announce it, it does not disappear on a timer, and it meets WCAG 2.2 AA contrast.

5. Marking AI-generated text

Article 50(2) requires providers of AI systems that generate text to mark outputs in a machine-readable way. We plan to watermark texts longer than 200 tokens. The watermark will be applied identically in “before” and “after” modes, so both produce comparable output and the speed-up stays independently verifiable.

Status: how this obligation applies to locally run models, and which method to use, is awaiting a legal opinion and a query to the Polish AI authority (KRiBSI). It is not implemented yet. We will describe it here before the App is released.

6. No images or video

Enigma LABZ does not generate images, audio or video, including deepfakes, so the Article 50(4) obligations for such content do not apply to us.

7. Our role under the AI Act

We are not a provider of a general-purpose AI (GPAI) model; the Models come from their creators. Our work concerns how a model is executed, and the compute used for any modification is far below the one-third-of-original-training-compute threshold that the European Commission’s guidelines use as the indicator for becoming a GPAI provider. We are, however, the provider of an AI system (the App) and we meet the Article 50 transparency obligations.

In Poland, the AI Act is supervised by the Commission for the Development and Safety of Artificial Intelligence (KRiBSI), established by the Polish Act on Artificial Intelligence Systems.

8. Reporting a problem

If a Model produced harmful or unlawful content, or you did not see the AI disclosure, write to us at the address below. With your permission, we will pass reports about a Model itself on to its creator.

Contact e-mail

— (to be completed after company registration)