AI Act notice
Last updated: 20 November 2026
This page sets out how Blixild AI relates to the EU AI Act: that its answers are produced by AI, what the models are for, what they must not be used for, and what falls to you when you use them. It describes the service; it is not a legal assessment of your own use of it.
1. Answers come from AI models
Every answer from the API at api.blixild.com, and from Run and Chat in the dashboard, is produced automatically by an AI model. No person writes or checks it. The models are open-weight models made by others, which Blixild runs in the EU.
2. What the models are for
Chat models, such as Bielik, write, summarise, translate, pull data out of text, answer from documents you provide, and work as the language model in your own tools and agents. Decision models (basal) answer a question you set about a text with a probability for each allowed answer, for example to route a message, sort a ticket or screen a document.
How well a model does depends on the model, the language and the task. Test it on your own data before you rely on it.
3. Limits
A model can be wrong, out of date, incomplete or biased, and still sound certain. The same question can get different answers. A model knows nothing of today's events or of your business unless you give it that information. A decision model's probability is an estimate: you decide at what probability your system acts on its own and when a person looks first.
Answers on legal, medical, financial, employment, education and safety matters must be checked by a competent person before anyone acts on them. Test the models for your purpose, watch for errors, and keep a way to switch the automation off.
4. Uses that are not allowed
Don't use the service:
- for any practice Article 5 of the AI Act prohibits. Among them:
- manipulating people, or exploiting their age, disability or social or economic situation, in ways that cause them significant harm;
- social scoring;
- predicting crime from profiling alone;
- building face-recognition databases by scraping images;
- inferring emotions at work or in schools;
- sorting people by biometric data into race, religion, political opinions or sexual orientation;
- from 2 December 2026, generating realistic intimate or sexual images, video or audio of an identifiable person without their consent, or child sexual abuse material;
- to decide about a person on a model's answer alone in an area the AI Act lists as high-risk (its Annex III): biometrics, critical infrastructure, education and training, employment and work, credit scoring, life and health insurance, public benefits and emergency services, law enforcement, migration and border control, and justice and democratic processes such as elections. Such a use needs its own legal assessment, documentation and human oversight, which the service doesn't provide.
Section 5 of the Terms of Service makes these rules part of the contract.
5. When you build on the API or publish its answers
When you build our models into your own product, you decide what it does and how its answers are used. The AI Act's duties for that product then fall on you, and which ones depends on your role and on what you build. Two that often apply:
- a product that talks with people, such as a chatbot, must tell them they are dealing with an AI, unless that is obvious (Article 50(1));
- text published to inform the public on matters of public interest must be disclosed as AI-generated, unless a person has reviewed or edited it and someone holds editorial responsibility for the publication (Article 50(4)).
6. Your data
We don't keep your requests or the answers after the answer is sent, and we don't use them to train models. The Privacy Policy has the details.
7. Reporting a problem
If a model gives a harmful answer, or you see the service used against these rules, write to contact@blixild.com. If you can, give the time of the request, the model and the name of the API key, which may help us find it, and leave out personal data we don't need. We aim to look into every report.
8. The law this page refers to
Regulation (EU) 2024/1689 (the AI Act), Articles 5 and 50 and Annex III, as amended by Regulation (EU) 2026/1744 (the Digital Omnibus on AI). The duties that apply to you depend on your role and on how you use the service.