Sherlock — detect unknown AI usage
Detect, measure, and monitor shadow-AI usage across your networks. Sample scan shown — all data below is canned.
Est. unknown AI spend
$342.18/mo
2,137,000 est. tokens
Devices using AI
14
93 conversations
Networks monitored
3
1 ignored
Providers detected
5
Top: OpenAI
Where unknown AI usage is coming from
Network locations by public IP · sized by detections
San Francisco, US
HQ-WiFi · 10.0.4.0/24
$210.40
58 detections
Austin, US
Branch-Guest · 192.168.1.0/24
$88.10
27 detections
Denver, US
VPN-Remote · 172.16.0.0/16
—
ignored
By provider
41 conversations · 128.4 MB
22 conversations · 61.2 MB
15 conversations · 38.9 MB
9 conversations · 19.4 MB
6 conversations · 12.1 MB
Networks
Costs are attributed per network. Ignore or delete any you don't want counted.
HQ-WiFi
San Francisco, US · 10.0.4.0/24
Branch-Guest
Austin, US · 192.168.1.0/24
VPN-Remote (ignored)
Denver, US · 172.16.0.0/16
Detected AI conversations
Per device → endpoint. Metadata only — Sherlock never reads message contents.
api.openai.com · 4.2 MB · ~180k tokens · 97% conf.
chatgpt.com · 6.8 MB · ~292k tokens · 99% conf.
api.anthropic.com · 2.8 MB · ~121k tokens · 95% conf.
generativelanguage.googleapis.com · 1.9 MB · ~84k tokens · 92% conf.
api.perplexity.ai · 940.0 KB · ~41k tokens · 90% conf.
api.deepseek.com · 610.0 KB · ~26k tokens · 88% conf.
Find shadow AI on your own network in minutes
This is the real interface — with your own keys it runs on live models, your data, and your team. Start a free 14-day trial (no credit card) or estimate your cost first.