> ## Documentation Index
> Fetch the complete documentation index at: https://help.honestly.com/llms.txt
> Use this file to discover all available pages before exploring further.

# How does Honestly comply with the EU AI Act?

> Which AI features Honestly has, where your data is processed and how to switch individual features off – a basis for your own assessment under the AI Act.

The EU AI Act regulates the use of artificial intelligence in the EU. This article shows where Honestly uses AI, who reviews the results, where the data is processed and how you switch AI features off for your account.

### Our classification

In our assessment, Honestly is a limited-risk AI system, not a high-risk AI system. Honestly makes no automated personnel decisions: the AI features analyze feedback and make suggestions, and decisions about people are always made by people.

Honestly is the provider of the AI system. Your company is the deployer as soon as it uses the AI features. This can bring obligations of its own, such as informing employees about the use of AI. Clarify them with your data protection officer or legal counsel.

### Which AI features there are

| Feature | What it does | How the result is used |
| - | - | - |
| **Topic AI for Open Comments** | Assigns open comments to topics | The assignment is saved and marked as *AI categorized* |
| **Sentiment AI for Open Comments & Bubble Chart** | Rates the sentiment of open comments | Feeds into the analysis on the dashboard |
| **AI-Powered Open Text Summarization** | Summarizes open comments | Appears as **AI summary of text responses** next to the original responses |
| **AI Initiative Recommendations** | Suggests initiatives that fit your results | Suggestions only – you choose what fits your team |
| **Data Masking** | Replaces personal data in open text responses with placeholders | Applied automatically; admins can undo individual maskings |
| **Speech-to-text** | Turns spoken answers into text while taking a survey | Participants see the text and send it themselves; the audio is not stored |
| **AI Analysis of this Data** in 360° feedback | Analyzes the results of a 360° feedback and writes an interpretation, development impetus and reflection questions | Suggestions – the person decides whether to act on them |
| **Company context** | On request, researches your company from its web domain and pre-fills the wizard | You review and edit every field before saving |

For more on data masking, see [Data masking](https://help.honestly.com/help/data-security/data-masking.md).

### Human oversight

No AI feature in Honestly makes decisions about people. Topics, sentiment and summaries are analysis aids: they are generated automatically, and the people reading the results decide what conclusions to draw from them. Initiative recommendations are explicitly suggestions that you adopt, adapt or discard.

### No emotion recognition

The AI Act prohibits AI systems that recognize emotions in the workplace. This means systems that infer emotions from biometric data, such as facial expressions or voice. Sentiment AI in Honestly only analyzes the written text of open comments. According to the European Commission's guidelines on prohibited AI practices, sentiment analysis of text does not fall under this prohibition. With speech-to-text, spoken words are only turned into text, and only that text is analyzed afterwards.

### Where the data is processed

* The AI features run on Microsoft Azure (Azure OpenAI, operated through Azure AI Foundry), and so does speech-to-text.
* The subprocessor is **Microsoft Deutschland GmbH**, with servers located exclusively in the EU.
* Connections are encrypted, and stored data is encrypted with 256-bit AES. Details are in the [IT security & data protection FAQ](https://help.honestly.com/help/data-security/it-security-faq.md).

### Switching AI features off

Surveys and dashboards work without the AI features. As an admin, you decide which of them are active in your account:

1. Open **Account Settings → Features**.
2. Click **Disable** next to the feature.

The feature is then hidden for all users in your account.

Under **Account Settings → Subprocessors** you also control whether Microsoft may be used as a subprocessor for your account. Consent is given by default. Without it, data masking, speech-to-text, **Automatic Topic Analysis**, the AI analysis in 360° feedback and the automatic research for the company context, among others, are not available.

> [!NOTE]
>
> Turning Microsoft off under **Subprocessors** does not on its own switch off every AI feature – for example, not AI-Powered Open Text Summarization or AI Initiative Recommendations. If you do not want to use AI features, also disable them one by one under **Features**.

### Documents for your works council and data protection

For your discussions with the works council and data protection officer, we are happy to provide the data processing agreement (DPA), the technical and organizational measures (TOMs), technical descriptions and the list of subprocessors, and to support you with the data protection impact assessment (DPIA). Write to [support@honestly.com](mailto:support@honestly.com).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.