AI policy
How OpX uses AI, and what it never does with your data.
AI features in OpX (AI Coach, SOP flowchart generation, digital-twin queries) make work faster and more insightful — without trading your data for it.
Three commitments
No training on customer data.
Customer data is never used to train shared AI/ML models. Our model providers operate under no-training contractual terms.
Anonymised prompts.
Where prompts would otherwise contain customer-identifying information, OpX anonymises before invocation.
Tenant isolation before AI.
Row-Level Security applies before any AI invocation. One tenant's data cannot appear in another tenant's prompt — by design.
AI sub-processors
| Provider | Use case | Data sent | Region |
|---|---|---|---|
| OpenAI | LLM inference (GPT-4o), embeddings (text-embedding-3-small) | Anonymised prompts, no training | US |
| ElevenLabs | Voice synthesis for learning content (where enabled) | Text content only | US |
| Jina AI | Web content extraction for research workflows | URLs only | EU |
- Provider
- OpenAI
- Use case
- LLM inference (GPT-4o), embeddings (text-embedding-3-small)
- Data sent
- Anonymised prompts, no training
- Region
- US
- Provider
- ElevenLabs
- Use case
- Voice synthesis for learning content (where enabled)
- Data sent
- Text content only
- Region
- US
- Provider
- Jina AI
- Use case
- Web content extraction for research workflows
- Data sent
- URLs only
- Region
- EU
Audit trail
Every AI invocation is logged in the per-tenant ai_invocation_log: who invoked it, what feature, when, and what completion was returned. Customers can audit their AI usage at any time.
Customer controls
Customers can disable AI-enabled features at the organisation level via configuration. AI features are always opt-in.
Regulatory monitoring
We monitor UK and EU AI regulation, including the EU AI Act, and adjust controls as regulatory clarity emerges. AI features are positioned as advisory — outputs are reviewed and approved by a human before any consequential action.