Overview
How might we design permission drift indicator so people can trust and act on AI output?
When this pattern fits
- Essential for AI agents, OAuth-based assistants, and connector ecosystems where long-lived permissions accumulate silently and users benefit from proactive visibility and review.
When to skip or lighten it
- Single-permission tools that never expand scope.
- Locked enterprise agents whose scopes are set only by admins.
- First-run onboarding before any drift could exist.
States
State model coming soon
Key UX elements
Key UX elements coming soon
Anti-patterns to avoid
Allowing scope expansion with no history of what changed.
Scary banners with no list of new permissions.
Drift alerts that cannot deep-link to revoke.
Comparing against averages that shame users without clear risk.
How products use it
| Product | Implementation |
|---|---|
| Google Account Security | Alerts and reviews for third-party access changes. |
| 1Password Watchtower | Surfaces risky or stale access patterns. |
| GitHub | Connected app reviews and excess permission warnings. |
| Apple Privacy Report | App access summaries that reveal unexpected usage. |
Implementation
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:Frequently asked questions
What is permission drift for AI agents?
Permission drift is when an agent collects more tools, data scopes, or integrations over time than the user originally understood or intended.
How should the indicator look?
A compact badge or banner listing new scopes since last review, with Review and Revoke actions. Avoid vague “security risk” copy alone.
How often should you prompt review?
On each scope expansion, and on a periodic cadence for always-on agents. Do not nag weekly if nothing changed.
How does this differ from scope disclosure?
Scope disclosure explains permissions at grant time. Drift indicators track how that set changed afterward.