AI Model Register
A complete register of third-party AI providers used by October Health, maintained in accordance with the EU AI Act and ISO 42001.
OpenAI
Models
Deployment contexts
Purpose within October Health
- Luna conversational AI (emotional support, check-ins)
- Ash coaching and L&D content generation
- HR document summarisation
- Assessment analysis
Data processing
| Property | Value |
|---|---|
| Training opt-out | Yes (contractual) |
| Conversation logging | Disabled |
| Data retention | 30 days |
| DPA in place | Yes |
| International transfers | Standard Contractual Clauses (SCCs) |
EU AI Act risk classification
Limited Risk (EU AI Act)
Hiring-support features using OpenAI models are classified High Risk and subject to additional human oversight.
Google Gemini
Models
Deployment contexts
Purpose within October Health
- Ivy AI Dietitian — nutritional advice and meal planning
- Multimodal document understanding
- Workplace Wellbeing Benchmark analysis
Data processing
| Property | Value |
|---|---|
| Training opt-out | Yes (contractual) |
| Conversation logging | Disabled |
| Data retention | Zero retention |
| DPA in place | Yes |
| International transfers | Standard Contractual Clauses (SCCs) |
EU AI Act risk classification
Limited Risk (EU AI Act)
Nutritional advice is presented as informational; clinical dietary decisions remain with registered professionals.
Anthropic
Models
Deployment contexts
Purpose within October Health
- Complex reasoning tasks within agent pipelines
- Safety-critical response generation (crisis detection layer)
- Long-context document analysis
- AI governance tooling (internal)
Data processing
| Property | Value |
|---|---|
| Training opt-out | Yes (contractual) |
| Conversation logging | Disabled |
| Data retention | Zero retention |
| DPA in place | Yes |
| International transfers | Standard Contractual Clauses (SCCs) |
EU AI Act risk classification
High Risk where used in crisis/safeguarding context; Limited Risk otherwise
Crisis detection outputs are always reviewed by a human before any action is taken.
October Health does not share user conversation content with any third-party AI provider for the purpose of model training. This is enforced contractually via Data Processing Agreements with each provider and verified during our quarterly audit cycle.
All providers listed above have confirmed in writing that October Health data is excluded from their training pipelines. We use API-tier access with zero or minimal data retention, ensuring that conversation content is not stored beyond what is necessary for immediate processing.