AI Agents security and privacy
Agents and their records stay inside your organization. Tool switches and an autonomy level are saved on each agent, ready for when agents can be deployed.
Tool switches
Tools such as web search, memory, code execution, email, messaging and calls to connected services are listed on the agent as switches. They are meant for a deployed agent, and deployment isn’t open yet, so they have no effect today.
Autonomy and approvals
Agents have the same autonomy levels as AI Staff: Suggest, Approve (the default) and Autonomous. They decide how much a deployed agent may do before asking a person. Until deployment opens, agents take no actions in the outside world, so the level has no practical effect.
Your data
- Agents belong to the organization that created them. Members of other organizations can’t see them.
- Removed agents can be restored for 30 days.
- The organization compliance settings include a switch for whether agent runs help improve Syntic agents. See Model training and your chats.
Learn more
Was this page helpful?