AI Act transparency notice

AI TRANSPARENCY · VERSION DATED 30 OCTOBER 2026

This page describes the origin of output, intended use and limitations of the Fabryka AI inference service. It is not a declaration of certification or a guarantee that the service or a user's particular use complies with the law.

1. Who you are interacting with

Responses from the Fabryka AI API are generated automatically by AI models. You are not talking to a person. The service is provided by Kacper Wikiel trading as Fabryka AI. The service may use different models, such as Bielik and Qwen. Bielik is a project developed by SpeakLeash and ACK Cyfronet AGH. CodeSOTA is a separate project for measuring and comparing models, rather than the name of this inference service.

2. AI origin of content

The service's output comes from an AI system. Information about that origin does not mean that a person has reviewed the output. Before using output, users should assess what verification their use requires. An AI-origin label alone does not confirm that output is correct, safe or lawful.

3. Intended use

The service is intended for general text generation and transformation, including summarisation, classification, extraction, drafting, RAG and automation. Capabilities, including the quality of Polish-language support, depend on the selected model and task. The service is not intended as a standalone system for making decisions with significant consequences for a person.

4. Limitations and human oversight

A model may hallucinate and provide incorrect, outdated, biased or incomplete information. Output may vary even with similar prompts. A model does not automatically know current facts or a user's private context.

Legal, medical, financial, employment-related, educational and safety-related content requires assessment by a competent person. Users should carry out tests appropriate to their purpose, monitor errors and ensure that automation can be stopped.

5. Excluded uses

The service must not be used for practices prohibited by Article 5 of the AI Act. Separate analysis, documentation or safeguards do not override those prohibitions; the law defines their scope and any specific exceptions.

Under the service's usage rules, without separate analysis, documentation and safeguards, the service must not be used as the sole basis for decisions concerning employment, credit, insurance, access to education or public services, justice, biometrics, healthcare or other high-risk areas. This service usage rule does not replace a legal assessment of a particular deployment.

6. Obligations when publishing output

Users should preserve any supplied AI-origin label. This is guidance on using the service; specific legal obligations depend on the use.

When a user publishes AI-generated or AI-manipulated text to inform the public on matters of public interest, the user is responsible for the required disclosure of its origin under Article 50 of the AI Act, subject to statutory exceptions. The exception concerning review or editorial control requires the content to have undergone human review or editorial control and a natural or legal person to hold editorial responsibility for publication. An API response label alone does not establish that these conditions have been met.

7. Models, data and version

The model list may change; the /v1/models endpoint provides the current list. The model and execution configuration may differ depending on the selected processing path. The service may use different computing resources depending on availability and the type of request.

The service terms describe the rules for using the service. The privacy policy describes data processing, categories of recipients and retention.

8. Risks and reports

Report an incident, harmful output or a compliance concern to k.wikiel@gmail.com, providing the request identifier if you have it and as little personal data as possible.

9. Basis of this notice

This notice refers in particular to Articles 4, 5 and 50 of Regulation (EU) 2024/1689 (the AI Act), as amended by Regulation (EU) 2026/1744. Obligations depend on the entity's role, the system's intended purpose and how it is used. This notice does not confirm fulfilment of organisational or technical obligations and does not replace a legal assessment of a particular deployment.