Refined Agent Parameters for Improved Safety and Control
The UnO Agentic AI Builder introduces refined agent parameters that provide greater control over agent behavior, safety, and execution stability.
Enhanced safety controls
Safety settings have been expanded with new safeguards to support responsible AI usage:
- Jailbreak Protection: Helps prevent attempts to override agent instructions or bypass intended behavior.
- PII Safeguards: Reduces the risk of sensitive personally identifiable information being exposed or mishandled during agent interactions.
Execution limits
New Agent Max Iteration settings allow you to define execution boundaries for agents and individual tool calls. These limits help prevent runaway loops, improve predictability, and optimize resource usage during agent execution.