# Trust & transparency

If You work with larger organisations or operate in a regulated industry, You may have been asked whether FitekAI complies with the EU AI Act. This page explains where FitekAI fits within the regulation, what it means for You as a customer, and why You can use it with confidence.

### What is the EU AI Act?

The EU AI Act is a regulation that classifies AI systems based on the level of risk they pose. It entered into force in 2024 and applies across all EU member states.

The regulation defines four levels of risk:

* **Unacceptable risk:** AI systems that are outright banned, such as social scoring or manipulative behavioural techniques
* **High risk:** AI systems used in sensitive areas like employment decisions, credit scoring, law enforcement, or critical infrastructure, which must meet strict requirements
* **Limited risk:** AI systems where transparency is the main obligation, meaning users need to know they're interacting with AI
* **Minimal risk:** AI systems that don't fall into any of the above categories, which are largely unregulated<br>

***

### Where does FitekAI fit?

FitekAI is an invoice processing tool. It suggests accounting dimensions, cost centres, and product codes based on Your company's own historical invoice data. It doesn't make decisions about people. It doesn't assess creditworthiness, score behaviour, process biometric data, or influence access to services. It helps You fill in transaction rows faster.

Under the EU AI Act, FitekAI falls into the **minimal risk** category. The high-risk use cases defined in the regulation - things like recruitment tools, credit scoring systems, law enforcement applications, and critical infrastructure management - don't apply to accounting automation.

That said, minimal risk doesn't mean no responsibility. There are still good practices that matter, and Fitek takes them seriously.

***

### Transparency and traceability

One of the core principles of the EU AI Act is that people should know when AI is involved and be able to understand what it did.

FitekAI is designed with this in mind:

* **Clear labelling.** Every change made by FitekAI is logged with the source tag "AI" in the transaction row history. You can always see what was filled in by AI, what came from system automation, and what was changed manually by a user.
* **No hidden decisions.** FitekAI proposes values - it doesn't execute transactions or make final decisions. Every suggestion can be reviewed, corrected, or overridden before the invoice moves forward in Your approval workflow.
* **Autotransactions stay in control.** If Your autotransaction rules cover the same fields, they take priority. FitekAI fills in the gaps, it doesn't compete with Your rules.

***

### Your data stays Yours

Each company gets its own AI model, trained exclusively on that company's historical invoice data. There is no cross-company data sharing. Your model learns from Your invoices, Your cost centres, and Your posting patterns - and nobody else's.

When You correct a suggestion, that correction feeds back into Your model. Over time, the suggestions become more accurate because they reflect how Your specific organisation handles its accounting.

***

### Human oversight

FitekAI doesn't remove people from the process. It handles the repetitive part - filling in dimensions and cost centres on transaction rows - so that the people reviewing invoices can focus on exceptions and approvals rather than data entry.

Every AI suggestion goes through the same approval workflow as any other invoice. Nothing is posted, exported, or finalised without human review. The people in Your organisation remain in control of every invoice at every step.

***

### AI literacy

Article 4 of the EU AI Act requires organisations to ensure that people working with AI systems have a sufficient understanding of how those systems work.

This documentation, along with the main FitekAI page, is designed to support that. It explains what FitekAI does, how it learns, what it proposes, and how its suggestions interact with other automation. If You need additional training or materials for Your team, contact Your Fitek account manager.

***

### Summary

| Topic               | How FitekAI handles it                                                                               |
| ------------------- | ---------------------------------------------------------------------------------------------------- |
| Risk classification | Minimal risk - not in any high-risk category defined by the EU AI Act                                |
| Transparency        | All AI-made changes are tagged with "AI" source in transaction history                               |
| Human oversight     | All suggestions are reviewable and overridable within Your existing approval workflow                |
| Data separation     | Each company has its own model, trained on its own data only                                         |
| Decision-making     | FitekAI proposes accounting values - it does not make decisions about people or execute transactions |
| Continuous learning | Manual corrections and finalised postings feed back into the model to improve accuracy               |


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://fitek-learning.fitek.com/fitek-learning-space/copy-of-english/fitekai/what-is-fitekai/trust-and-transparency.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.