Allow
The prompt proceeds. Logged at the category level your policy defines.
How it works
From prompt to decision—with policy citations built in.
A direct path: intercept on supported AI sites, evaluate against your active policy, enforce inline with guidance employees can act on. No content dragnet—four deterministic steps.
- 1
Intercept at submit
The browser extension captures prompts on supported AI sites—scoped to the tools your policy covers, not the whole browser.
- 2
Evaluate your policy
Your published AI use policy drives allow, warn, block, or log-only. Decisions return in seconds with the policy section cited.
- 3
Coach in the moment
Employees see clear guidance tied to your rules. Security gets structured signals—not silent bulk surveillance.
- 4
Learn in aggregate
Skill-extraction and nudge generation run strictly out-of-band, after the decision—coaching that never slows the 4-second before-send path.
Four outcomes, in your policy's language.
Every decision cites the exact standard the employee was held to—so audits trace to policy, not to a vendor's black box.
Warn with redaction
The employee sees what was detected and gets one-click redaction, or proceeds with a business justification.
Block
Nothing leaves the browser. The dialog explains which policy section applies and why.
Log only
For approved-context tools: the event is recorded for coaching patterns without interrupting anyone.
What happens before anything is collected
No content is ingested until an administrator attests, on the record, that everyone affected received written notice. The attestation is versioned and disclosed in plain language—including exactly how terminal-session buffering works. Employee notice templates for NY, CT, DE, and CO are built in.