Use AgentShield to reduce AI risk, not to ignore it.
These terms explain the practical expectations for using AgentShield during early access. They are written to be readable because trust should not hide behind dense legal copy.
Last updated: May 28, 2026
You control your rules and data
You decide which prompts, outputs, and rules you send to AgentShield. You are responsible for making sure your use of the product matches your own policies and user obligations.
AgentShield is a safety layer, not a guarantee
The product can help detect and block risky text, but no automated guardrail can promise perfect protection. Keep human review for high-risk actions.
Keep API keys private
API keys allow requests to be logged under your workspace. Do not expose them in public repositories, client-side code, screenshots, or support chats.
Early access can change
AgentShield is still improving. Features, limits, and pricing may change as the product matures, and we will aim to communicate important changes clearly.
Do not use AgentShield as the only approval system for dangerous actions
For refunds, account deletion, medical, legal, financial, or other high-impact decisions, keep a human approval step and your own application-level controls.
Questions before production?
We can help you think through a safe rollout path.