1. Data Protection Impact Assessment (DPIA)

Project: Mindora Pilot – AI Questionnaire & Avatar Intake
Date: 15.09.2025
Responsible Officer / DPO: Dr Bhaskar Punukollu

Summary

Mindora is piloting an AI assistant with 50 patients. The AI will:

— Transcribe a short (5-min) avatar interaction into text (non-identifiable ADHD-related examples only).
— Automate scoring and summarisation of results from pre-validated, licensed ADHD rating scales.
— Collate results into a structured summary for the clinician.
— Store all results securely in Semble EHR (system of record) for 6 months.
— Will not make inferences, provide diagnoses, or offer treatment suggestions.

Nature, scope, context, purpose

— Nature: Special category data (health-related) with anonymised/non-identifiable content.
— Scope: 50 patients in pilot.
— Context: Supportive tool for clinicians, not diagnostic.
— Purpose: Reduce administrative burden through transcription, collation, scoring.

Roles & responsibilities

Controller: Mindora / Dr Bhaskar Punukollu

Processors:
— OpenAI — AI transcription and collation (Business mode, no storage/training).
— Semble — secure storage of patient data.

Art. 28 Accountability: DPAs in place with both processors.

Lawful basis

— GDPR Art. 6(1)(e)/(f): Public interest or legitimate interest in healthcare.
— GDPR Art. 9(2)(h): Healthcare purposes under clinician oversight.
— Explicit patient consent obtained.

Data flows

— Input: Anonymised questionnaires + 5-min non-identifiable ADHD symptom examples.
— Processing: OpenAI Business mode (no storage/training; encrypted transit/at rest).
— Output: Stored only in Semble EHR for 6 months.

Risks and mitigations

— Identifiability: Low risk – mitigated by anonymisation and careful prompt design.
— Unauthorised access: Mitigated with MFA, RBAC, encryption.
— Processor misuse: Mitigated with Art. 28 DPAs and audits.
— Errors/bias: Mitigated by clinician oversight.
— Patient misunderstanding: Mitigated with consent form + privacy notice.

2. Patient Privacy Notice

Who we are:

Mindora is testing an AI assistant to support clinicians.

What we collect:

— Anonymised responses from ADHD questionnaires.
— A short 5-minute avatar interaction with non-identifiable behavioural examples.

Why we use it:

— To automate scoring and summarisation of licensed ADHD questionnaires.
— To reduce administrative workload.
— The AI will not make inferences, diagnoses, or treatment suggestions.

Where your data is stored:

— In Semble EHR, GDPR compliant.
— Retained for 6 months.

Who processes your data:

— Mindora/Clinic (Controller).
— OpenAI (Processor): Business mode, no storage/training, encrypted.
— Semble (Processor): Secure storage.

Your rights:

— Access, correction, deletion, restriction.
— Withdraw consent at any time.

Contact:

DPO: Dr Bhaskar Punukollu, dr.bpunukollu@mindora.ai
Date: 15.09.2025

3. Internal AI Governance Policy

Responsible Officer: Dr Bhaskar Punukollu
Date: 15.09.2025

Scope and Limitations:

— AI use is limited to transcription, collation, and scoring of licensed ADHD scales.
— No inferences, diagnoses, or treatment suggestions are generated.
— Only anonymised, non-identifiable content is processed.

Processor Oversight:

— OpenAI: DPA ensures no storage/training, encrypted processing.
— Semble: DPA ensures compliant storage and 6-month retention.

Governance:

— AI Lead: operational oversight.
— DPO: compliance & audits.
— Clinician oversight of all outputs.

Incident Response:

— Breaches reported within 72h to ICO if required.
— Patients notified if risk is high.

4. Data Mapping Register (GDPR Art. 30)

Intake transcription

Purpose: Summarise for clinician
Data Categories: Non-identifiable ADHD examples
Controller: Clinic (Dr Punukollu)
Processor(s): OpenAI (Business mode)
Storage/Retention: Semble, 6m
Legal Basis: Art. 9(2)(h), consent

Questionnaire collation

Purpose: Score/summarise licensed ADHD scales
Data Categories: Anonymised questionnaire data
Controller: Clinic
Processor(s): OpenAI (Business mode)
Storage/Retention: Semble, 6m
Legal Basis: Art. 9(2)(h), consent

Output report

Purpose: Provide to clinician
Data Categories: Derived scores, summaries
Controller: Clinic
Processor(s): Semble
Storage/Retention: Semble, 6m
Legal Basis: Art. 9(2)(h)

5. Access Control and Security Policy

Responsible Officer: Dr Bhaskar Punukollu
Date: 15.09.2025

Principles:

— Least privilege, MFA, encryption (TLS 1.2/1.3, AES-256).
— Access logged for audit.
— Annual staff training.

Processor Security:

— OpenAI Business mode: anonymised, non-identifiable content only, no storage/training.
— Semble: GDPR-compliant, secure hosting.

Responsibilities:

— IT Lead enforces controls.
— Access revoked on staff exit/change.

6. Scope of AI Use Statement

Project Lead / DPO: Dr Bhaskar Punukollu
Date: 15.09.2025

What the AI system does:

— Transcribes a short 5-minute avatar interaction with non-identifiable ADHD-related examples.
— Processes anonymised responses from licensed ADHD questionnaires.
— Automates scoring and summarisation of questionnaire results.
— Stores summaries securely in Semble EHR for 6 months.
— Processes data in OpenAI Business mode (no storage/training, encrypted).

What the AI system does not:

— Does not process patient identifiable information.
— Does not provide diagnoses.
— Does not make treatment suggestions.
— Does not generate inferences beyond transcription and scoring.
— Does not replace clinical judgment.
— Does not use data for training, profiling, or secondary research.

Accountability and Oversight:

— Controller: Mindora / Dr Bhaskar Punukollu.
— Processors: OpenAI (Business mode), Semble EHR.
— Legal basis: Explicit patient consent + GDPR Art. 9(2)(h).
— Clinicians oversee all outputs; AI acts only as administrative assistant.