Training & HR
Skills development, feedback, and onboarding tools. Ethical AI integrated by default — no ethical rules to configure.
ProfessionalsAI in your organisation.
AI in your organisation.
Ethics in the code, not in your procedures.
You deploy AI tools in your domain. ELYSÉA CERVEAU integrates into the pipeline — not into your methods. Ethical constraints are in the SDK, non-deactivatable, with an open measurement protocol.
Integration sectors
Consulting & advisory
Structuring, restitution, and synthesis assistants. The ELYSÉA pipeline is designed to enforce the 4 relational prohibitions at every call that goes through it — manipulation and domination targeted by the pipeline, not only by a prompt.
Public services
Citizen-facing applications, digital counters, guidance portals. Active prohibitions protect your users without additional audits.
Regulated sectors
Finance, education, healthcare (outside clinical acts). Prohibition traceability is inscribed in the pipeline — not in a revisable internal policy.
What integration brings
AI Act points addressed in the pipeline
Transparency, anti-manipulation and redirection in critical zones are enforced by the pipeline, not only in your terms of service. They help meet the AI Act without amounting to compliance on their own: the compliance analysis remains yours.
Zero ethical configuration to maintain
Ethical constraints are in the SDK. Your technical teams do not implement or test them — they are non-deactivatable by construction.
Verifiable label for your users
The "Powered by ELYSÉA" label is linked in real time to the SDK access status. If access is revoked, the key is refused immediately, the SDK label shows “Not verified” and the withdrawal is entered in the public register; the builder removes the label within 48 hours. It is an architectural proof, not a contractual mention.
AI Act Compliance
What the law requires.
What ELYSÉA inscribes in the pipeline.
The European AI regulation imposes legal obligations on AI providers in professional contexts. The ELYSÉA pipeline handles several of these points without configuration; they cannot be disabled through configuration after integration. Your compliance assessment and audits remain yours.
Art. 5 AI Act
Prohibition of manipulation
The AI Act prohibits AI systems that exploit biases or manipulate behaviour against user interests. ELYSÉA Prohibition P5: no oriented bias mechanism, no hidden persuasion — designed to apply at every call that goes through the pipeline.
Art. 50 AI Act
Transparency & identification
AI systems in direct interaction must identify themselves as AI. ELYSÉA Prohibition P4: the AI never claims to be human, identifies itself as an AI pipeline at every relevant interaction — non-deactivatable by the builder.
AI Act § education
Prohibition of emotional evaluation
AI cannot evaluate emotional states in educational and professional contexts. The ELYSÉA Academy configuration blocks this capability via runtime posture — inscribed in the canon, non-configurable by the school.
GDPR + AI Act
Traceability & user rights
Every pipeline refusal is traced and timestamped; the log is provided on request. Your users can consult, export and delete their memory at any time — GDPR rights inscribed in the pipeline, not in a revisable internal policy.
What ELYSÉA is designed never to do
These prohibitions are in the pipeline — non-deactivatable by the application builder, non-negotiable by contract. Designed to:
- No diagnosis — regardless of how the user phrases their message
- No prescription — no treatment, designed to give no directive medical advice
- No unconsented influence — no oriented bias, no hidden persuasion
- No human pretence — the AI identifies itself as an AI pipeline at every relevant interaction
- No algorithmic retention — no engagement mechanics, no behavioural optimisation
- Mandatory redirection in critical zones — urgency, distress, qualified need: orientation towards a competent human
- Documented human procedure for crisis triggers: ELYSÉA directs, follow-up is yours.
Frequently asked questions
- Is ELYSÉA compatible with our existing practices?
- Yes. ELYSÉA integrates into the technical pipeline — it does not modify your method, your interface, or your relationship with your users. The prohibitions apply downstream of your application logic.
- Will your users see ELYSÉA?
- Yes. The "Powered by ELYSÉA" label is mandatory for every integrator — linked to the SDK access status, it cannot be removed without revoking that access. It is a commitment to your users, not a presentation option.
- How long does integration take?
- SDK integration is documented and designed to be fast. Timelines depend on your stack. Full documentation is available at /en/sdk — start with the quickstart.
Ready to integrate ELYSÉA into your tools?
Your dossier
Dossier 02 — Entreprises qui équipent leurs salariés
- Problème : émotions au travail interdites (art. 5(1)(f)) ; CSE à consulter (L2312-8) ; formation (art. 4 remplacé) [S-DRT-01] [S-DRT-04] [S-DRT-02].
- ELYSÉA : verrou émotions, interdits, trace pour l'assureur, orientation en crise.
- Réserves : région Supabase eu-central-1 (Francfort), relevé du tableau de bord par Sébastien le 01/10 à 18:03, pièce M-I1 à joindre ; chemins vers des prestataires hors UE possible, en cours de retrait — repli OpenAI, mémoire vectorielle, voix (OpenAI TTS, Deepgram ; retrait du cerveau décidé le 01/10 à 11:27) ; réserve maintenue jusqu'au déploiement vérifié ; DPA non signé.
- Offre : Équipes 30 / 18 / 9,60 € par siège et par an (
elysea-web:src/lib/publicFacts.ts:49-56@00c9c89). - Prochain pas : consultation du CSE. →
02_ENTREPRISES_EQUIPANT_SALARIES.md
Measurements (campaign results file)
- Over-blocking
- Response quality