Permission Gate
A mechanism in AI agent workflows that requires human confirmation before an action is executed. It is characterized as ex ante, binary, and instantaneous, and tends to produce rubber-stamp approvals and automation bias, thereby dispersing rather than reducing risk.
Timeline (last 90 days)
The rate at which agents complied with injected instructions barely changed with the gate design, with 13 to 16 compliance events per 300 attacks.
The key difference between accountability and permission gates lies in the temporal dimension: permission gates are ex ante, binary and instantaneous; accountability is ex post, continuous and accumulative.
All Facts (2)
The rate at which agents complied with injected instructions barely changed with the gate design, with 13 to 16 compliance events per 300 attacks.
50%UnverifiedThe key difference between accountability and permission gates lies in the temporal dimension: permission gates are ex ante, binary and instantaneous; accountability is ex post, continuous and accumulative.
50%