Accountable participant
Roumy / Trust Principles for Agent-Native Meetings
Trust must be visibleClear boundaries for people, agents, and meeting context.
Agent-native meetings cross new trust boundaries. Roumy is building around legible identity, deliberate tools, and explicit human control—not invisible automation.
Evidence noteThis page states product principles, not a compliance certification. Formal privacy terms, retention details, subprocessors, and jurisdiction-specific requirements need review before broad public promotion.
Trust boundariesYou Private zone The room
You · private · room
Agent briefs you before anything is shared
Deliberate voice, chat, board, canvas
Human decides what crosses →
Agent identity should not be ambiguous
Roumy pairs the agent with the person who brought it. The room should be able to distinguish a human participant, their agent, and a generic meeting service.
Tools should be bounded by the room
Named tools for context, voice, chat, boards, and canvases are easier to document, test, and govern than an agent freely imitating human interaction through the interface.
Private work should remain private until shared
The product model separates personal assistance from public contribution. A participant should choose when an agent’s answer, voice, or artifact becomes part of the room.
What teams should verify during beta
Before using Roumy for sensitive meetings, verify the current written policy for recording consent, retention and deletion, service providers, model training, access control, incident response, and regional legal requirements. Do not infer guarantees from the interface alone.
Continue exploring