AI Transparency

1. AI System Overview

NEURIDION uses artificial intelligence to classify Field Safety Notices (FSNs) as part of post-market surveillance (PMS) for medical device manufacturers. The AI serves as a classification tool that analyses publicly available regulatory notices and determines their relevance to your specific device profile.

AI does not make clinical decisions, diagnose patients, recommend treatments, or replace the judgment of a Person Responsible for Regulatory Compliance (PRRC). All AI outputs are advisory and require human review before acting on them.

2. Models Used

Anthropic's API usage policy states that inputs sent via the API are not used to train their models.

3. How AI Classification Works

NEURIDION uses a two-stage classification pipeline:

1. Pre-filter (Haiku):Each FSN is evaluated for obvious irrelevance to your device profile. Approximately 60–70% of FSNs are clear exclusions (e.g., a dental device FSN when you manufacture cardiac implants) and are handled at this stage for speed and efficiency.
2. Full classification (Sonnet): Remaining FSNs receive a detailed analysis. Each is classified as relevant, uncertain, or excluded, accompanied by a written rationale and a confidence score between 0.0 and 1.0.

4. Human Oversight Measures

• Items classified as "uncertain" are explicitly flagged for PRRC review — they are never silently excluded.
• If the AI filter fails for a specific FSN, that item is marked as "requires manual review" rather than being silently dropped.
• Users can cancel any running search at any time.
• Generated reports include a signature grid for PRRC sign-off, ensuring a qualified person reviews and approves the results.
• All filter decisions are stored in an immutable, append-only audit trail — decisions cannot be edited or deleted after the fact.

5. Data Handling

What is sent to AI

• FSN title, manufacturer name, date, and content text
• Device profile context (device name, manufacturer, intended use, device class)

All data sent to the AI consists of publicly available regulatory notices published by government agencies (BfArM, FDA, MHRA, Swissmedic) and device profile information you have entered.

What is NOT sent to AI

• Your email address, password, or authentication credentials
• Patient data or protected health information (PHI)
• Health or medical records of any kind
• Payment information or billing details
• Internal company documents or proprietary files

6. Known Limitations

• All AI classifications are advisory only and must be reviewed by a qualified person before being relied upon for regulatory decisions.
• Confidence scores are model estimates, not calibrated statistical probabilities.
• Novel or rare device types may receive less accurate classifications due to limited representation in training data.
• Classification accuracy is highest for English and German language FSNs; other languages may yield reduced accuracy.
• The AI does not have access to proprietary device documentation, IFUs, or internal risk assessments — classification is based solely on the FSN text and device profile.
• All AI-generated outputs must be reviewed and approved by the PRRC before inclusion in official PMS documentation.

7. Prohibited Practices Statement (EU AI Act Art. 5)

NEURIDION does not engage in any practices prohibited under Article 5 of the EU AI Act. Specifically, the system does not:

• Deploy subliminal, manipulative, or deceptive techniques
• Exploit vulnerabilities related to age, disability, or social/economic situation
• Evaluate or classify individuals based on social behaviour (social scoring)
• Perform real-time remote biometric identification

The system's sole function is classifying publicly available government-published Field Safety Notices against user-defined device profiles.

8. EU AI Act Risk Classification

Under the EU AI Act (Regulation (EU) 2024/1689), NEURIDION's FSN classification system is assessed as minimal risk. The system classifies publicly available government-published regulatory notices — it does not process biometric data, make decisions affecting natural persons' rights, or fall within any Annex III high-risk category. A Data Protection Impact Assessment (DPIA) screening has been conducted and is available upon request.

9. Contact

For questions about our use of AI, data handling, or this transparency disclosure, contact us at info@neuridion.eu.