Smart Approvals in Amazon Quick: Human oversight where it matters

Amazon added Smart Approvals to Quick, a review layer that automatically approves eligible agent actions or returns them to the user, so security leaders can keep human attention for decisions that need it.

• Permissions still set the boundary: denied actions stay denied, and mandatory human approvals still apply, with an action eligible only when the user's setting is "Ask each time" and no other control requires approval.
• A separate classifier model judges each eligible action against the user's request and a bounded view of the conversation, weighing what the action does, where data goes and who it affects, and can consult the user's knowledge graph, which stays within the Quick account.
• Quick says it does not use the content the classifier reviews, including knowledge graph data, to train models or improve the service.
• Each automatic approval is single-use, applies only to the exact action and parameters reviewed, and grants no ongoing permission; actions the classifier does not approve, timeouts and other review failures return to the user.
• Administrators control who can use it through Custom Permissions, where it is on by default but profiles with deny-by-default must explicitly allow it, and users can turn it off for themselves across conversations and scheduled runs.
• The permission service records permission updates associated with Smart Approvals in AWS CloudTrail, and Quick also records permission-check verdicts and changes to a user's Smart Approvals setting.