Permission drift indicator is an AI interface design pattern that proactively reminds users which permissions their AI agents currently hold, when those permissions were granted, and when they were last reviewed. This UX pattern combats the slow, invisible accumulation of scope that happens across months of OAuth grants, connector installs, and "just this once" approvals: the failure mode where users wake up to discover an assistant has read access to three email accounts, write access to two drives, and billing access to an account they barely remember authorizing. The indicator surfaces drift on a visible cadence (weekly digest, inline nudge, settings badge), makes each granted capability auditable, and provides one-click revoke. It is the audit-log pattern for the agent era: transparency as preventative security.
Essential for AI agents, OAuth-based assistants, and connector ecosystems where long-lived permissions accumulate silently and users benefit from proactive visibility and review.
Copy this prompt to generate a production-ready implementation in Cursor, Claude Code, Lovable, or any AI coding agent.
Generate a production-ready implementation of the "Permission Drift Indicator" AI interface design pattern.
Pattern Definition:Settings → Permissions
1 permission not reviewed lately.
Read inbox
Allowed 12 days ago
Last reviewed: 2 days ago
Send email
Allowed 93 days ago
Last reviewed: never
View calendar
Allowed 34 days ago
Last reviewed: 14 days ago
Collapsible thinking and tool traces
Hover for source
Show how sure the model is about a claim
Concrete actions to validate uncertain output
Visualizing RAG
Attach verifiable sources to generated claims