What it means
Risk tiering is sorting AI systems, or the actions an agent can take, into levels of risk so the controls match the level. Low-risk actions can run with light checks, while higher-risk actions get stronger controls such as human approval, tighter limits, closer logging, or a ban. Tiering starts from what an action could affect — money, data, safety, or people's rights — and how hard it would be to undo.
Why it matters for governance
It matters for governance because it focuses effort where harm is possible, instead of treating every agent action as equally safe or equally dangerous.
Example
A team tiers an agent so drafting an internal summary is low-risk, emailing a customer needs review, and changing a payment is blocked without human approval.
Related ActLattice products
General information only — not legal advice.