Orygent AI and Identity Use Policy
1. Scope
1.1 This policy applies to every AI feature of Orygent: role-based assistants, department agents, drafting, analysis, workflow proposals and any generated content. It forms part of the Member Account Terms (ACC-01) and of the Customer's subscription terms (CUS-03 / ENT-01).
1.2 Orygent does not offer voice cloning, likeness synthesis or biometric identification. Section 5 states what applies if such a feature is enabled by a later registry version.
2. AI disclosure, outputs and escalation
2.1 Disclosure. Where you interact with an AI system (chat, voice, agent, digital twin or generated content), the Operator tells you so at the start of the interaction and labels AI-generated or AI-manipulated content in accordance with Article 50 of Regulation (EU) 2024/1689. A voice or persona representing a real person is used only with that person's recorded permission.
2.2 Nature of outputs. AI outputs are generated from models and data that can be incomplete or wrong. They are information and assistance, not professional advice, and not a decision of the Operator. You remain responsible for decisions you take on the basis of an output. Where a decision has legal, financial, medical, educational or safety effect, the service routes it to a human reviewer or to you for decision, and the Operator does not make that decision automatically.
2.3 Human escalation. You may at any time request a human review of an AI interaction through support@orygent.com or the in-product escalation control; the Operator responds within 5 business days (consumer) or within the SLA (business).
2.4 Permitted use. You will not use AI features to impersonate a person without permission, to generate unlawful content, to circumvent safety controls, or to extract or reverse-engineer models. The Operator may suspend AI features for breach.
2.5 Recording and model use. Recordings, transcripts and uploads are used to provide the service to you. They are used to improve models only with your separate permission (purpose research_contribution), which you may withdraw at any time without affecting the service.
3. Enterprise AI operations
3.1 The customer appoints tenant administrators who manage users, authority levels and data classes within the tenant. Actions with legal, financial or employment effect require approval by a human with the configured authority before execution; the service does not execute them automatically.
3.2 Employee data processed in the tenant is processed under the customer's instructions as controller; the DPA (DAT-01) applies.
3.3 Outputs are recommendations and drafts; decisions are the customer's.
3.4 Every assistant identifies itself as AI, names the Tenant it serves and shows the sources and evidence labels attached to its output. Missing facts and conflicts between sources are shown, not resolved silently.
3.5 An assistant acts only with the tools, data classes and tenant scope the tenant administrator has enabled for the role. Tool access is revocable by the administrator and every tool call is recorded in the audit system.
3.6 An assistant does not use one Member's private preferences or another Tenant's content to answer a different Member.
3.7 Recruitment, employment, legal, financial and customer-binding actions proposed by an assistant are placed in the approval queue of the person holding the configured authority. The approver's decision, identity and time are recorded. An approval cannot be delegated to an AI.
4. Training and confidentiality
4.1 Tenant content, conversations and uploads are not used to train third-party models.
4.2 The Operator uses Tenant content to improve its own models only where the Customer has granted the research_contribution permission in tenant settings; the permission is off by default and withdrawal takes effect for future processing.
4.3 Operator staff do not read confidential Tenant content; support access is limited to operational, log and technical records and is audited (PUB-02 clause 8.2).
5. Voice, likeness and recording (not enabled)
5.1 Orygent does not capture or synthesise voice or likeness and does not perform biometric identification. If a later registry version enables such a feature, it is governed by a separate explicit permission (recording_voice or likeness_persona) collected before first use, states whose voice or image is used, the audience, purpose, retention and withdrawal route, and is never inferred from a profile photo or account registration.
6. Reporting unsafe output
6.1 Report an unsafe, unlawful or misleading output through the in-product report control or support@orygent.com. The Operator acknowledges within 2 business days and a person reviews the report. In an emergency, contact the local emergency services; Orygent is not a monitoring or emergency service.
7. Changes
7.1 The Operator may update this policy under ACC-01 clause 9. Each version is archived with its date and content hash at the versioned archive route of this document.