AI AGENT DISCLAIMER
This page explains what the AI agents in Atom9 can and cannot promise, what activity is recorded and why, and what changes when you connect an AI agent of your own.
AI output can be wrong. Check important results before you rely on them. Every AI action in your workspace is recorded so problems can be found, fixed, and — if it ever matters — attributed. If you connect your own agent, its behaviour is your responsibility, and the same recording applies.
AI models generate text and actions from patterns. They can be confidently wrong, miss context, or misunderstand what you meant. Treat AI answers the way you would treat a capable colleague's first draft: useful, fast, and worth checking before anything important depends on it.
Anything with real consequences in Atom9 — publishing, approvals, changes that affect your customers — asks for your explicit decision first. The agent can prepare work; you approve it.
Where a workflow touches something regulation cares about, the agent is required to stop and ask before acting. That is a platform rule, not a model preference.
Each AI conversation, tool call, and system command in your workspace is recorded with a correlation ID and kept in your workspace's compliance area. This includes what the agent was asked, what it did, and what the system changed as a result.
Three reasons. First, debugging: when a build goes wrong, the record shows exactly which step failed. Second, security: unusual or unsafe agent behaviour can be detected and investigated. Third, accountability: if there is ever a dispute about who or what did something — a person or an agent, yours or ours — the record answers it.
The record lives in your workspace and is visible to your workspace admins. Atom9 operations staff can access it when investigating an incident, a support request you raised, or suspected abuse — access that is itself logged.
The managed A9 agent runs on infrastructure Atom9 operates. If a workspace configures an AI provider whose processing happens outside the EU/EEA, the console discloses that next to the chat before you type — that disclosure is generated from the deployment's configuration and cannot be turned off.
If you connect an AI agent or model of your own to your workspace, you are responsible for what it does there — its outputs, its instructions, and the provider terms behind it. Atom9 provides the rails: permissions, approval gates, and tenancy boundaries apply to your agent exactly as they do to ours.
All AI commands are monitored regardless of whose agent issues them. Your agent's actions are recorded in the same compliance area, with the same correlation IDs, for the same reasons: debugging, security, and settling who did what.
The chat always shows which agent is selected, and the disclaimer under the chat changes when you switch to an agent of your own — so there is never ambiguity about whose model you are talking to.
For anything this page does not answer — including what is stored about your workspace's AI activity and how to request a copy — use the privacy request form or see the privacy policy.
Privacy
You can change this later from the footer. Read the Cookie Policy and Privacy Policy.