PENGO Workbench
Workbench / AI-Adjacent Issues / Permission Change Not Applied

AI-Adjacent Issue

Permission Change Not Applied

A permission, approval, role, policy, or access change is made but the AI, agent, tool, or workflow continues behaving as if the old permission state still applies.

What This Looks Like

A permission, approval, role, access rule, policy, or workspace setting changes, but the AI, agent, tool, or workflow continues acting as if the old state still applies. A revoked permission may still appear active, a new permission may not take effect, or a changed role may not alter behavior.

Why Users Blame AI

The behavior appears in the AI workflow, so the user may think the AI ignored the permission change. The underlying cause may be cached state, delayed propagation, connector state, account permissions, role mapping, stale policy context, or a runtime that has not received the updated authority state.

What to Check First

  • Whether the permission change was applied in the correct account, workspace, or environment
  • Whether the connector or tool needs reauthorization
  • Whether the runtime is using cached permission state
  • Whether role changes apply to the agent, tool, and workflow layers
  • Whether revoked approval is still present in prompt or workflow context
  • Whether docs and runtime behavior reflect the same permission model

When This Is AI-Adjacent

Use this AI-Adjacent Issue when a permission or authority change appears not to apply at runtime. If a revoked approval continues to govern behavior, use the related Workbench Issue for revoked approval still treated as active. If declared policy and actual behavior differ, use that related Issue.

Related Workbench Issues

Common Ways People Describe This

  • permission change not applied
  • AI still uses old permission
  • permission update not reflected
  • role change not applied to agent
  • approval change not active
  • access change not reflected in AI workflow