> For the complete documentation index, see [llms.txt](https://docs.ruler.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ruler.nl/about-ruler/ai.md).

# AI at Ruler

Ruler's position on AI today, the commitments that apply to AI functionality in development, and how Ruler relates to the EU AI Act.

## Where Ruler stands today

Ruler contains **no AI functionality in production**. Content is collected by scrapers and assessed, processed and published by Ruler's legal editorial team, as described in [How Ruler's content is produced](/about-ruler/editorial-process.md). Ruler does not use AI to decide what is relevant to you, to summarise regulation or to propose changes to your compliance framework.

A first AI feature is in development. This page describes the commitments that apply to it, so you know in advance how Ruler will handle AI once it is introduced. Where a statement below concerns that feature rather than the current product, it says so.

## Commitments for AI functionality

Ruler falls under the same AI governance framework as CERRIX: a group-level AI policy, and an assessment of every AI application before it goes live. The framework is described in detail on the [CERRIX AI](https://docs.cerrix.com/about-cerrix/trust-and-compliance/cerrix-ai) page. For Ruler, it comes down to the following.

**Human in the loop.** AI output in Ruler will be presented as such and will never change your compliance or control framework without a person in your organisation deciding to apply it. There is no automatic effect without human intervention.

**Traceable to the source.** AI output will be recognisable as AI-generated and will always be traceable to the original regulatory source, in the same way that every alert and radar item is today.

**No training on your data.** Customer data and user interactions are never used to train or improve models.

**Processing within the EU.** AI processing takes place within the European Union.

**Models and provider.** The AI feature in development uses OpenAI GPT foundation models, accessed through Microsoft Foundry (formerly Azure AI Foundry). Ruler works with pinned model versions in Microsoft Foundry: there is no automatic upgrade to a new model version.

## How model changes are managed

A change of model, model version or prompt logic follows Ruler's regular change management process: a change request, review, acceptance testing and formal approval before anything reaches production. This process is part of the SOC 2 control framework of Ruler B.V. (see [Trust and compliance](/about-ruler/trust-and-compliance.md)).

How changes are communicated:

* Functional changes are announced through the [release notes](/help/release-notes.md) and Ruler's release communication to customers.
* Changes that materially affect the behaviour or outcomes of AI functionality, and any change of model provider, are announced in advance.
* A change of sub-processor follows the notification procedure in the data processing agreement.

## EU AI Act

Because Ruler has no AI functionality in production today, no obligations under the EU AI Act currently rest on the product.

The AI feature in development has been classified in advance. It does not fall under the high-risk category of the AI Act and qualifies as limited risk. Ruler meets the transparency obligation that applies to it: AI output is recognisable as such and always traceable to the original source, and there is no automatic effect without human intervention. AI literacy is safeguarded through the CERRIX AI policy and the associated training.

{% hint style="info" %}
Questions about AI at Ruler that this page does not answer? E-mail Ruler support at <info@ruler.nl> or get in touch with your Ruler contact. We are happy to walk your compliance, legal or security team through the setup.
{% endhint %}