> For the complete documentation index, see [llms.txt](https://doc.dastra.eu/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.dastra.eu/en/features/ai-systems/establishing-a-record-of-ai-systems.md).

# Establishing a record of AI systems

To set up a record of your AI systems, you have two options:

* Either you already have a registry that you can import directly into Dastra (go to the next section: [Import your AI systems](/en/features/ai-systems/import-your-ai-systems.md))
* Or you don't have one. In this case, you'll have to create one yourself

## Create a record of AI Systems

To add an AI system, click **"Create a new AI system"**. A window offers three creation modes:

<figure><img src="/files/zd1ozJ6hljqSvdhyPzyc" alt="Create a new AI model dialog with three options: Generate with AI, Built-in template, Personalised"><figcaption><p>Three creation modes: AI generation, built-in template, or blank form</p></figcaption></figure>

* **Generate with AI** — The AI assistant automatically generates a complete record from a short description (name, publisher, URL…). It uses web browsing to enrich the information.
* **Built-in template** — Choose from the Dastra AI system library or from another workspace.
* **Personalised** — Create a blank record and fill in each field manually.

### Generate an AI system with the AI assistant

Select **"Generate with AI"**, then describe the system to document (name, publisher…). You can attach a file or provide a URL to help the generation.

<figure><img src="/files/Oki0VezZEMOqh6sOetxY" alt="AI system generation form with description field and URL option"><figcaption><p>Describe the system to generate — the assistant uses web browsing to complete the information</p></figcaption></figure>

The assistant produces a pre-filled record (name, description, learning type, systemic risk, publisher…) which you can review and validate before saving.

<figure><img src="/files/gaP1OIw3Isdf0xU8vDxX" alt="AI-generated record for the Gemini 3.1 Pro system"><figcaption><p>The generated record should be reviewed before validation — AI can make mistakes</p></figcaption></figure>

{% hint style="warning" %}
Data generated automatically by AI may contain errors. Review and correct the proposed record before saving it to your registry.
{% endhint %}

Once you've entered the required information, you'll be redirected to a 10-step form. This form will enable you to give as much detail as possible about the AI system.

<figure><img src="/files/24o4Gbnv8qFQgmQAefrD" alt=""><figcaption></figcaption></figure>

## The 11 Steps of the AI System Form

Below are the 11 steps to complete when documenting an AI system.

***

### 1. General

Enter **basic information** about the AI system:

* **Name** of the system
* **Brief description** of its purpose and functionality

***

### 2. Responsibilities

Define your **role and responsibilities** under the European AI Act:

* **Provider**: develops and places the AI system on the market
* **Deployer**: uses the AI system within professional activities

***

### 3. AI Models

Specify the **AI model(s)** used to process data within this system.

> ℹ️ For more details, refer to the AI Models Repository.

***

### 4. Stakeholders

Identify **stakeholders involved** in implementing and managing this AI system, including their roles (e.g. Data Scientist, DPO, Product Owner).

***

### 5. Assets

Add the **assets supporting this AI system**, such as:

* Infrastructure components
* Software tools
* APIs
* Documentation resources

***

### 6. Datasets

List the **datasets associated** with this AI system. Indicate their usage among the following phases:

* **Training**: the dataset used to **train the AI model**, enabling it to learn patterns, relationships, or classifications based on historical data.
* **Validation**: a separate dataset used to **tune model parameters and prevent overfitting**. It helps assess model performance during training and guides adjustments for optimal results.
* **Testing**: another distinct dataset used to **evaluate the final performance** of the trained and validated model before deployment. It provides an unbiased measure of how the model will perform on new, unseen data.
* **Production inference**: data processed by the AI system **during actual operation**, where the trained model generates predictions, classifications, or decisions in real-world scenarios.

***

Ensure that each dataset’s **purpose, composition, and linkage to this AI system** are clearly documented for transparency and compliance purposes.

***

### 7. Data Subjects

Specify the **categories of data subjects** whose personal data is processed by the AI system (e.g. customers, employees, users).

***

### 8. Risk Analysis

Assess the **level of risk** based on:

* Types of data processed
* Processing activities
* Potential impacts on individuals’ rights and freedoms

***

### 9. Business Value

Determine a **business value score** reflecting the system’s contribution to your organization to:

* **Prioritize** high-value projects
* Align AI initiatives with strategic objectives

***

### 10. Documentation

Attach relevant **documents and information leaflets**, such as:

* User notices
* Technical guides
* Compliance assessments (e.g. DPIAs)

***

### 11. Summary

Review a **comprehensive summary** of all information entered for this AI system before final validation and registration.

***

## Linked processing activities and data synchronisation

The **Data processings** tab of an AI system record lets you associate one or more processing activities from your GDPR record with that system. This link keeps your processing register and your AI systems registry consistent with each other.

<figure><img src="/files/QXQ9XVEv2912XX3PaMAW" alt="Data processings tab in the AI system record with the Sync. data button"><figcaption><p>The "Data processings" tab lists the GDPR processing activities linked to this AI system</p></figcaption></figure>

### Syncing data from a processing activity

For each linked processing activity, the **"Sync. data"** button automatically imports relevant elements into the AI system record: assets, datasets, and data subject categories.

<figure><img src="/files/zGfkauwrDOvwXYWx6SOv" alt="Sync. data button highlighted in the linked processing activities list"><figcaption><p>The "Sync. data" button triggers synchronisation from the processing activity into the AI system record</p></figcaption></figure>

A dialog lists the elements available for import, letting you select exactly which ones to add.

<figure><img src="/files/WYHOfY5JUHEGUvDDsqeh" alt="Synchronisation dialog showing assets, datasets and subject categories to import"><figcaption><p>Select which elements to sync from the processing activity into the AI system</p></figcaption></figure>

### Viewing linked AI systems from a processing activity

Conversely, from any processing activity record, a **"Linked AI systems"** side panel lists all AI systems associated with it.

<figure><img src="/files/S74FtMx2ds3YxjipiIkE" alt="Linked AI systems panel in a processing activity record"><figcaption><p>The "Linked AI systems" side panel is accessible from any processing activity record</p></figcaption></figure>

This two-way link simplifies audits and AI Act compliance documentation from within the GDPR processing register.

***

[Risk analysis and business value](/en/features/ai-systems/risk-analysis-and-business-value.md)\
\
[Import your AI Systems](/en/features/ai-systems/import-your-ai-systems.md)

[Transparency notice](/en/features/ai-systems/transparency-notice.md)

[AI Models repository](/en/features/ai-systems/ai-models-repository.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## 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, and the optional `goal` query parameter:

```
GET https://doc.dastra.eu/en/features/ai-systems/establishing-a-record-of-ai-systems.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

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.