AI Model Register
Last updated: April 2026
This register is maintained in accordance with Section 2 of our AI & Data Policy. It documents the AI models in use within the SENSA platform, their intended function, input and output data types, and the safeguards applied to each.
Schools, MAT data leads, and DPOs may request additional detail by contacting contact@sensatech.co.uk.
Current Status
SENSA is currently evaluating third-party foundation models for production deployment. Providers under evaluation include:
- Google (Gemini)
- OpenAI (GPT series)
- Anthropic (Claude)
No provider will begin processing school data until they have been formally approved, added to our sub-processor list, and schools have been notified in accordance with our 30-day advance notification commitment.
Model Usage by Platform Function
Once finalised, the following table will be updated with specific model identifiers and versions.
| Platform Function | AI Role | Input Data | Output Data | Model Provider |
|---|---|---|---|---|
| SENSA Observe | Pattern-recognition analysis of anonymised classroom work signals to generate a ranked need-signal watchlist | Pseudonymised pupil identifiers, classroom work patterns, need-type signal data | Ranked watchlist with probability-weighted need indicators for SENCO review | Under evaluation |
| SENSA Adapt | Content transformation into parallel adaptations for six SEN need-type profiles | Curriculum content and need-type instruction only (no pupil-identifying data) | Adapted lesson materials optimised for dyslexia, ADHD, autism, SLCN, dyscalculia, EAL+SEN | Under evaluation |
| SENSA Evidence | Structured data compilation and formatting of statutory-format EHCP documentation | Intervention records, need-signal history, teacher notes (pseudonymised) | Draft EHCP evidence packs for SENCO review and approval | Under evaluation |
Safeguards Applied to All Models
| Safeguard | Implementation |
|---|---|
| No PII in prompts | Technical controls prevent transmission of personally identifiable pupil data to any external AI API. Enforced at the application layer. |
| Pseudonymisation | Pupil Data is pseudonymised at ingestion. Real identifiers are stored separately and cannot be reconnected within the AI processing pipeline. |
| Human-in-the-loop | All AI outputs are recommendations for professional review. No automated decisions with legal or similarly significant effects on children. |
| Bias monitoring | Models reviewed for bias and fairness at least annually against representative UK pupil demographics. Quarterly output auditing by advisory SENCO panel. |
| No training on school data | SENSA does not permit AI model providers to use school data for model training or improvement. This is enforced contractually with each provider. |
| Data residency | Where available, UK or EEA-based inference endpoints are used. International transfer safeguards are documented on the sub-processors page. |
Change Notification
Schools will be notified at least 30 days in advance of any change to:
- The AI models used to process Pupil Data
- The nature of data inputs to AI models
- The sub-processors used for AI inference
- Any feature that materially changes how need-signal generation operates
Schools that object to a proposed change may terminate their subscription without penalty by providing written notice within 14 days of the change notification.
Transparency Reporting
SENSA will publish an annual AI Transparency Report documenting the results of bias and fairness reviews. The first report will be published following the completion of the initial production deployment.
Contact
For questions about our AI models, to request a copy of the full AI Model Register, or to request our Data Processing Agreement, contact contact@sensatech.co.uk.