Never decide for a public body
ClariDroits does not replace the French bodies MDPH, CDAPH, CPAM or CAF, a doctor, a legal professional or a social worker. Decisions remain with the competent bodies and professionals.
Responsible AI for disability, health and social support
The proposed principles intended to guide ClariDroits' domain-specific AI.
The ClariDroits project intends to use AI to help people read, understand, organise and prepare a case file. AI can make mistakes, omit information or produce invented information, so every result would need to be reviewed and checked before use.
Human English clarity reviewed — accessibility testing pending — ClariDroits is currently a project led by Patrick François Kahouadji, before the legal entity is established.
The page language is British English (en-GB), the reference country is France and the jurisdictional context described is French. These are separate: choosing English does not select UK law or activate any French right, procedure, deadline or public availability.
Review status: Human English clarity was reviewed on 30 July 2026. Patrick approved this informational English web release. Native British English review and human accessibility validation are not claimed; no accessibility-conformance claim is made.
Human control
Support understanding without taking over decisions.
ClariDroits is intended to support understanding and preparation. It must not decide, diagnose, promise an entitlement or replace professionals.
Explain the reasoning
The tool must show the elements it has identified: available documents, evidence that may need strengthening, points requiring care, criteria to clarify and cautious next steps.
- Show why a point has been raised
- Keep supporting information visible
- Make outputs reviewable
Keep the user in control
Texts, summaries, drafts and checklists are intended to remain editable, readable and verifiable. The controls actually available over what is kept, shared or sent must be checked for every surface and version.
Protect sensitive information
The project favours a local-first approach and states that sensitive information must not be sent automatically to a third party, public body or external service without explicit consent. Actual data flows, storage and consent must be audited for every surface and version.
Caution and traceability
Use verifiable sources and refuse promises of outcomes.
Public information must favour official resources, verifiable reference points and careful wording.
Refer to sources where possible
When a rule may change, the content must invite the user to check it with the competent body.
- Official resources
- Verifiable reference points
- Cautious wording
Refuse promises
The declared principle is not to display a probability of approval, a chance score, a promise of reimbursement, guaranteed support or wording that suggests an outcome is certain. Its application must be checked for every version.
Respect lived reality
The domain-specific AI should help connect medical, functional, social and legal information with everyday life, including mobility, fatigue, housing, carers, employment, isolation, social participation and concrete evidence.
Prepare, do not replace
ClariDroits aims to help prepare a question, summary, appointment, case file or cautious appeal. Results may contain errors and do not replace human assessment or professional support.
Language safeguards
Wording that ClariDroits must not use.
These statements would wrongly suggest that a right or outcome had already been established.
- You are entitled to this support.
- Your application will be accepted.
- Reimbursement is guaranteed.
- Support will be provided automatically.
- This decision is certain.
- You will win your appeal.
Careful explanations
Wording that keeps uncertainty and human review visible.
ClariDroits should identify what may be useful to check without predicting a decision.
- This point may be worth checking with the competent body.
- This evidence may help document the situation more clearly.
- This option may be worth exploring in light of your actual circumstances.
- A professional review may be useful before anything is sent.
- ClariDroits helps organise information without predicting a decision.
Publication safeguards
Limits that remain visible in every language
- ClariDroits is intended to support understanding and preparation; it must not decide, diagnose or promise an entitlement.
- Professionals and competent public bodies retain their roles and decisions.
- Sensitive information must not be sent automatically to a third party without explicit consent.
- Actual safeguards, data flows, storage and consent must be checked for every surface and version.
- Rules that may change must be checked against current official French sources.
- French bodies and procedures are identified as French and are not replaced with supposed UK equivalents.
- The English-language page does not provide UK legal, medical or administrative guidance.
Key facts
Clear answers about this page
- What does responsible AI mean for ClariDroits?
- It means proposing that outputs remain cautious, explainable, reviewable and under human control while rejecting automated decisions and promises of outcomes. Actual effectiveness still requires verification.
- Does ClariDroits decide for a public body?
- No. Decisions remain with the competent bodies and professionals.
- Does this English page adapt French disability procedures to the UK?
- No. It explains the French project in English and provides no UK administrative pathway or legal equivalence.
Frequently asked questions
Questions about this page
Does ClariDroits make decisions for a public body?
No. ClariDroits is intended to help people understand and prepare a case file, without deciding for a French MDPH, CPAM or CAF, a doctor, a legal professional or another professional.
Why does ClariDroits set out responsible-AI principles?
Disability, health and social-support procedures concern sensitive situations. The project therefore proposes explainability, caution, human review and control as objectives whose effectiveness still needs to be verified.
Does ClariDroits use chance scores?
The declared principle is not to use chance scores, success probabilities or wording that could suggest a decision has already been secured. Its application must be checked for every version.