BriefDashboard
AI Discovery Report
● 24-hour window/Auto-refresh on
You're viewing sample data. This is a live demo workspace pre-seeded with realistic incidents, tools, and policies.
Get started · 1 of 2
Finish setting up Parker Industries
Two quick steps. Each one ticks the moment that surface phones home. You'll see real numbers on this dashboard the moment your first detection lands.
✓
Install the browser extension
Browser detections are landing. Roll the extension out to the rest of the team when you're ready.
2
Invite your colleaguesRECOMMENDED
Per-user view across the company. Each teammate gets their own signed audit log. You can also run Northbeams solo.
Daily brief
Generated Jan 6, 08:30 AMHIGH91%
What needs your decision
Two tools are leaking regulated data and should be blocked today. DeepSeek Chat has 24 users, all on personal accounts, with no DPA on file. In the last 24 hours it took pasted source code, employee SSNs, and new-hire home addresses. ChatGPT personal has 96 users and just took AWS keys and a Postgres connection string from Engineering. My recommendation: block both at the gateway and notify the 120 affected users, then redirect them to ChatGPT Enterprise and Claude org, which are already sanctioned and cover the same work.
What changed this week
Perplexity climbed to 51 users, up from 38 last week, and is now your most-used unsanctioned tool after ChatGPT personal. Three new tools showed up that I have not seen here before: Gamma (22 users), tl;dv (9), and Loom AI (33). All three handle meeting recordings or decks that routinely carry customer names. They are unclassified and sitting in your decision queue.
Exposure
Estimated data exposure is $670K across 143 sensitive prompts in the last 24 hours, driven mostly by the four high-risk tools. The deterministic policy engine already blocked the two worst pastes (source code into DeepSeek, credentials into ChatGPT personal) before they left the device. The residual risk is the long tail of personal-account usage that no policy covers yet.
Compliance
Your EU AI Act tool inventory is about 80% classified: 5 of 22 tools still need a risk decision. Three tools that process customer-facing content (Notion AI, Grammarly, Gamma) need a transparency notice before they meet Article 50. Identity provider, endpoint coverage, and sensitive-content monitoring are all green. The acceptable-use policy is drafted but not yet enforced.
By the numbersThe figures behind the brief
Recent activity
Last 24 hours
TimeEmployeeAI toolWhat was leakedAction
14:21:00
9m ago
MJPasted source code into DeepSeek Chat↳ 412 lines from billing-service/auth.ts. Detected: API keys, source code.14:06:00
24m ago
SCPasted customer data into Perplexity↳ 1,204 contacts and contract values from Acme-Q2-pipeline.csv.13:43:00
47m ago
ROPasted contract / legal into Claude (personal)↳ NDA redline mentioning codename and counterparty.+ 13 more events → view allShowing 3 of 16
Industry comparison
You're carrying more shadow than your peers
Your team 20 tools
20
Industry avg (SaaS · 1k-5k employees)18 tools
18
Top quartile (well-governed) 9 tools
9
11% higher
You're carrying 2 more AI tools than companies your size - most of it on the unsanctioned side.
See the 11 you could retire →© 2026 Northbeams IncStatus · Workspace dashboard · v1.0