AI Transparency Statement
Last updated: March 2026
1. Purpose and Scope
This statement explains how SEN Tools, provided by The Anthill Group (ICO Registration: ZB945587), uses artificial intelligence to assist education professionals in analysing Education, Health and Care Plans (EHCPs) and related SEN documentation. It is intended to support Data Protection Officers, Information Governance leads, and senior leadership teams in understanding the technical architecture that protects pupil data.
2. How AI Is Used
SEN Tools uses generative AI, accessed via secure API connections, to process anonymised document text and produce structured outputs. The AI is used to:
- Generate SMART plans from EHCP outcomes
- Create Support Snapshots summarising behavioural, medical, and safety needs
- Produce pupil profiles (About Me) from strengths and Section A content
- Organise instructional and environmental strategies (Strategy Builder)
- Summarise provision requirements (Provision Summary)
All outputs are advisory and intended as a starting point for professional review. They are not a substitute for professional judgement.
3. AI Providers
SEN Tools uses API-based AI services from providers such as OpenAI and Anthropic. These are commercial API services, not consumer-facing products like ChatGPT. Our architecture is model-agnostic — we can utilise the best available AI engine without any change to the data protection position, because the specific provider is irrelevant to the school's risk profile. This is explained in the next section.
4. The Anonymisation Bridge
The Anonymisation Bridge is a server-side middleware layer that sits between SEN Tools and any external AI provider. It ensures that no personally identifiable information (PII) ever leaves our UK infrastructure.
How It Works
- Detection and Tokenisation: Before any text is sent to an AI provider, the Anonymisation Bridge scans the document and replaces all PII with neutral tokens. For example:
- “Jayden” becomes “[NAME_1]”
- “Mrs Thompson” becomes “[NAME_2]”
- “15/03/2018” becomes “[DOB_1]”
- “SW1A 1AA” becomes “[POSTCODE_1]”
- NHS numbers, UPNs, emails, and phone numbers are similarly tokenised
- AI Processing: The AI provider receives only the tokenised text. It processes the request and returns a response that still contains only tokens — never real data.
- Restoration: Our server restores real values from a temporary in-memory vault. This vault is never logged or persisted to disk and is destroyed immediately after de-anonymisation.
Why This Matters
Traditional AI tools rely on contractual guarantees — a Data Processing Agreement with the AI provider that promises not to misuse your data. SEN Tools takes a different approach: the AI provider physically cannot access personal data because it never receives it. This is an architectural constraint built into how the system works, not a policy promise.
5. Categories of PII Protected
The Anonymisation Bridge detects and tokenises the following categories:
| Category | Examples | Token Format |
|---|---|---|
| Names | Children, staff, family members | [NAME_1], [NAME_2] |
| Dates of birth | DD/MM/YYYY, YYYY-MM-DD | [DOB_1] |
| UK Postcodes | SW1A 1AA, M1 1AA | [POSTCODE_1] |
| Phone numbers | 07xxx xxxxxx, +44 7xxx | [PHONE_1] |
| Email addresses | user@example.co.uk | [EMAIL_1] |
| NHS numbers | 3-3-4 digit pattern | [NHS_1] |
| Unique Pupil Numbers | Letter + 12 digits | [UPN_1] |
6. No AI Training
SEN Tools uses API-based AI services that do not train on, or learn from, your submitted content. Your data is used solely to generate the output you have requested. Additionally, because of the Anonymisation Bridge, even the anonymised tokens sent to AI providers contain no real personal information.
7. Human Oversight
All AI-generated outputs from SEN Tools are advisory only. They are designed to save time and improve clarity, but they are not a substitute for professional expertise. Users must always review, verify, and adapt generated outputs for the individual child's needs before use. SEN Tools does not make decisions about children — professionals do.
8. Accuracy and Limitations
AI-generated content may:
- Contain errors or omissions
- Misinterpret ambiguous text in source documents
- Produce outputs that require professional adjustment
- Vary in quality depending on the clarity and completeness of the source document
Users are responsible for verifying all outputs against the original documentation and applying their professional judgement before relying on any AI-generated content.
9. Audit Logging
SEN Tools maintains audit logs that record tool usage for accountability purposes. These logs record which tool was used, when it was used, and the type of processing performed. Audit logs do not contain any personally identifiable information about the children whose documents are processed.
10. DfE Generative AI Product Safety Standards
SEN Tools is designed to align with the Department for Education's “Generative AI: Product Safety Standards” (updated January 2026). In particular:
- The intended purpose and educational use cases are clearly stated
- The platform complies with UK GDPR and the Data Protection Act 2018
- Data processing, sub-processors, and data residency are fully transparent
- Pupil data is never used to train AI models
- Audit logging is maintained for accountability
- The Anonymisation Bridge ensures the tool operates as a “closed” system with respect to personal data
11. Contact
For further information about our AI architecture or to request a detailed technical briefing, please contact:
The Anthill Group
ICO Registration: ZB945587
Email: privacy@theanthill.co.uk
Address: 13a Park Street, Minehead, Somerset, United Kingdom, TA24 5NQ