Platform · 02 · Privilege Watch

Every detection, in one queue for legal.

Name: Privilege AI
Brand: Privilege AI

Privilege Watch is the console that collects detections from Claude, ChatGPT, and the Privilege AI Agent. It identifies the questions that need legal protection, routes them to counsel, and reports to legal — while everyday AI use carries on untouched.

Book a demo Next: Privilege AI Agent

Privilege Watch: Agent detections with open, new, needs-context and escalated counts, a 14-day detection activity chart, and top categories

Detail · top categorieslatest 6

Three sources. One review queue.

Detections are review suggestions, not confirmed violations. Legal decides what becomes a privileged session, what needs context, and what closes.

01 · Claude and ChatGPT

Conversation alerts from the company's AI accounts.

Watch detects the use of AI tools like Claude and ChatGPT by employees on company accounts, and flags the conversations that carry a legal question — a redline, a disciplinary letter, a bid comparison.

02 · Privilege AI Agent

Workstation detections across 120+ categories.

Everything the endpoint agent observes lands here, mapped to the four queues — Security, Privacy, Employment, Legal — with the category and the evidence that supported it.

03 · Routing and reporting

What needs protection goes to counsel. The rest is reported.

A flagged question opens a Privilege AI session with counsel present. Everything else rolls into the weekly report for legal: detections by queue, category, and status. Everyday AI use is not affected.

In the productAgent detections · Claude alerts · Agent setup · Claude setup

Privilege Watch summary tiles: 6 open detections, 6 new, 0 needs context, 0 escalated — Queue and activityOpen, new, needs context, escalated — and the 14-day trend.Counts exclude closed reviews and false positives. Detection counts do not measure device coverage; the heartbeat does.

Detection activity: 6 received in the last 14 days, by day — Queue and activityOpen, new, needs context, escalated — and the 14-day trend.Counts exclude closed reviews and false positives. Detection counts do not measure device coverage; the heartbeat does.

Privilege Watch, Agent setup tab: download the Mac agent, enrolled Macs with coverage and last-seen, and the organization capture policy — Setup lives in WatchAgent setup and Claude setup, beside the queues they feed.Enroll devices, check coverage, choose detection categories, connect Claude Enterprise — without granting device administration any access to legal content.

The counsel inbox: sessions awaiting counsel, counsel requests, and the basis review backlog that routed questions land in — Routed to counselA flagged question becomes a session in counsel's inbox.The requester keeps working. Counsel sees it waiting, joins the session, and the answer is recorded on the matter with its basis.

Top categories in the latest detections: privacy rights, external AI legal advice, contracts, employment decisions, sensitive-data egress — Top categoriesTwenty categories, four queues, one view of where the risk is.A detection can match more than one category. The report to legal is cut the same way, so the queue and the memo agree.

Specification

Detection of AI use that needs legal protection.

Provisioned by legal and security. Read by legal.

SourcesClaude Enterprise · ChatGPT · Privilege AI Agent

IdentifiesQuestions needing legal protection · 20 categories · 4 queues

Review statesNew · needs context · escalated · closed · false positive

Routes toA Privilege AI session with counsel present

Reports toLegal · weekly · by queue, category, status

Everyday useNot affected · detections are review suggestions, not verdicts

Available onWeb · Windows · Mac · iOS · Android

What a week of detections looks like

Real workflows, by department. Each one is a question legal would rather see before discovery does.

Sales · ChatGPTCustomer MSA redline pasted into a chatNeeds counsel → Privilege AI session opened on the customer matter.

HR · ClaudeDisciplinary letter drafted in a chatbotEmployment queue → routed to counsel before it is sent.

Engineering · AgentSource code pasted into a personal AI accountSecurity queue · sensitive-data egress → alert to legal and compliance.

Procurement · ChatGPTSupplier bids compared in a consumer toolLegal queue · procurement and third-party obligations → review.

Marketing · AgentCustomer identifiers uploaded to a new AI vendorPrivacy queue → DPA review before processing continues.

Marketing · ClaudeBlog outlineEveryday use · not affected · not reported.

United States v. Heppner · AI & legal privilege

Recognize the legal question before it becomes routine AI work.

Employees may turn to AI before they realize a question belongs with legal. Watch brings supported AI alerts and Agent detections into a review queue so counsel can assess the context and decide what should move into a matter. Earlier routing supports a better process for subsequent work; it cannot reverse an earlier disclosure.

See why early legal involvement matters

One Legal Risk Governance Platform

01 · Privilege AIThe privileged workspace

Where a routed question becomes a privileged session: counsel present, citations recorded, work product protected.

03 · Privilege AI AgentEDR sensor for compliance

The workstation sensor whose detections fill the Agent queue: 120+ categories, alerted to legal and compliance.