AI TRANSPARENCY
What RYTHM AI does, what data it can use, and where human authority remains mandatory.
Effective 14 August 2026. RYTHM Company OS uses clearly identified AI Agents and AI-assisted workflows for analysis, drafting, meetings, recommendations, and structured operational support. AI output is not presented as an undisclosed human decision.
WHAT IS AI
AI participation is explicit.
- Named AI Agents can generate analyses, recommendations, plans, drafts, and meeting contributions.
- Meeting summarization and governed deliberation can use AI to process authorized meeting context.
- Legal-review features are issue-spotting and escalation aids, not autonomous legal determinations.
- The public Demo is synthetic and read-only and is designed to demonstrate AI-company workflows without real customer production actions.
HUMAN AUTHORITY
Consequential authority remains with humans.
- The Human CEO remains the final authority for consequential company decisions in the current Public Beta governance model.
- AI Agents operate within defined roles, authority levels, risk ceilings, and approval boundaries.
- Recommendations are not silently converted into consequential external actions.
- AI output should be reviewed by an appropriately qualified human when business, legal, financial, safety, employment, or other significant consequences are involved.
MODEL DATA
Only relevant authorized context should be sent for inference.
- An AI invocation can include the user's prompt, agent role/instructions, meeting purpose, agenda, decision question, relevant transcript excerpts, and other authorized workspace context needed for the requested task.
- Current audited Public Beta AI inference uses OpenAI as the model provider when an AI feature is invoked.
- RYTHM should not intentionally send passwords, session tokens, API keys, service-role credentials, unrelated personal data, or another tenant's information to a model.
- Special-category, highly sensitive, or separately regulated data requires deployment-specific review before intentional AI processing.
PRIVACY
AI processing follows the same privacy and tenant boundaries as the service.
- OpenAI is disclosed in the current Subprocessor Register.
- The Privacy Policy explains AI-related personal-data processing.
- Where RYTHM acts as a processor for customer personal data, the DPA applies unless a signed agreement replaces it.
- RYTHM does not claim zero provider retention, EEA-only model processing, or provider no-training guarantees unless the applicable contracted configuration has been verified.
NOT APPROVED BY DEFAULT
Some AI uses require a separate legal, privacy, and risk review.
- Autonomous hiring, firing, promotion, worker scoring or worker-management decisions.
- Creditworthiness, insurance, education-access, essential-service, medical, biometric, law-enforcement, migration, justice, or safety-critical determinations.
- Manipulative, deceptive, exploitative, social-scoring, or other prohibited practices under applicable law.
- Any workflow that removes required human review from a consequential decision.
LIMITATIONS
AI output can be wrong.
- Generative AI can produce incomplete, inaccurate, outdated, fabricated, or contextually unsuitable output.
- Confidence, fluency, or formatting does not establish factual correctness.
- Customers are responsible for selecting appropriate human reviewers and lawful use cases.
- Security, privacy, model, and regulatory controls are reviewed as the Public Beta evolves.
AI GOVERNANCE REVIEW
Planning a regulated or consequential AI workflow?
Request a deployment-specific review before enabling the workflow in Production.