AI & LLM Policy

This policy describes how Hilma (operated by Mindtrack AB) uses Artificial Intelligence and Large Language Models (LLMs) to provide our services.

1. LLM Providers

Hilma uses the following LLM providers to power AI features:

2. Data Tenancy Policy

Hilma's LLM operations are processed as follows:

• OpenAI: API requests are processed in OpenAI's infrastructure in the United States. Hilma uses the standard API tier with zero data retention.
• Anthropic: API requests are processed in Anthropic's infrastructure in the United States. Hilma uses the standard API tier with zero data retention.
• No shared tenancy: Each API request is processed independently. Your data is not shared with other customers or used to train models.

3. Data Residency Policy

Hilma's LLM data residency:

• Processing location: LLM inference occurs in the United States via OpenAI and Anthropic APIs.
• Data at rest: Customer Data is stored in our primary database in the EU (AWS eu-central-2, Zurich). LLM providers do not store Customer Data.
• Transient processing: Data sent to LLM providers is processed transiently and is not persisted by the provider.

4. LLM Retention Settings

Hilma's LLM providers are configured with zero retention:

• OpenAI: Zero data retention. API inputs and outputs are not stored or used for model training. OpenAI's data usage policy for API customers confirms that data submitted via the API is not used to train models.
• Anthropic: Zero data retention. API inputs and outputs are not stored or used for model training per Anthropic's commercial API terms.
• No logging by providers: Neither provider logs or retains prompts, completions, or any Customer Data submitted through the API.

5. Data Sent to LLMs

The following types of Customer Data may be sent to LLM providers for processing:

• Check-in entries (daily recaps, weekly plans) for summary generation
• Retrospective notes for grouping and action item suggestions
• Journal entries when using AI Coach features
• Wellness response context for personalized coaching
• User messages when interacting with AI Coach

Note: Personally identifiable information (names, email addresses) is minimized in LLM prompts where possible. User IDs and workspace IDs are used instead of names in most contexts.

6. AI Output Storage

AI-generated content is stored in Hilma's database:

• Summaries: AI-generated team summaries are stored for the life of your subscription.
• AI Coach responses: Conversation history is stored to provide context for future interactions.
• Retro insights: AI-generated groupings and action items are stored with the retrospective session.

7. Opting Out

AI features in Hilma are integral to the service (summaries, coaching, insights). If you have concerns about AI processing, please contact us at [email protected] to discuss your requirements.

8. Contact

For questions about our AI and LLM practices, contact us at [email protected].