AI use at work

Know where company data goes when employees use AI.

Openwake shows which AI tools people use at work, by person and machine, labels the kind of data sent to each one, and shows what each provider says it does with it.

  • 5 endpoints free
  • No card required
  • No network changes

No keystrokes. No screenshots. No prompt content kept by default. The kind of data is worked out on the machine.

Openwake Search people, apps…Last 30 days
What needs your attention.
6open detections · 2 high
6policy violations
12systems reporting
50%browser coverage
Recent activity All events Policy violations
Oct 16Nov 14
Live151 vendors indexed1349 source documents53 sources unavailable627 processor relationships408 changes detected, 30 daysevery fact quoted with its source
Enforce policy · try it

You have an AI policy. How are you enforcing it?

In Chrome, Openwake checks what is about to go to an AI site. It can block an API key or a private key, and warn about customer data or source code sent from a personal account. Pick something to send, then change the mode or the account and try again.

1 · Pick something to send
2 · Your organization's mode
Which AI account

In this demo you choose the account.

Personal AI
‹ › ⟳🔒chat.personal-ai.comO
Pick one above, or type your own message.

After each send, this says what happened and what the admin sees.

A simulation in your browser. Nothing you type leaves this page. The real control works in Chrome only. Only API keys and private keys are ever blocked. AKIAIOSFODNN7EXAMPLE is AWS's documentation example, not a real key.

The problem

Your employees are already using AI. The hard part is knowing how.

A domain list shows where traffic went. Openwake shows who sent what kind of data to which provider, and what that provider's own documents say about it.

  1. Your teamPerson
  2. AI toolsApplication
  3. ProvidersProvider
  4. ProcessorsKnown processors
Sensitive data can travel further than you expect.
The Openwake model

Declared. Observed.

What the provider publicly promises and what is actually happening on your machines, in one record. The gap between the two is what you write policy about.

Declared

What providers say they do, in their own documentation.

  • Data usage
  • Training
  • Subprocessors
Registry

Observed

What actually happens, in your environment.

  • AI tools in use
  • Providers really reached
  • Kinds of data sent
  • Policy violations
Scout
Example from the demo

One detection, and why it was raised.

Openwake turns activity and classification into one record a reviewer can act on. This one is from the demo organization, as it stands right now.

MediumCredential-like indicator detected in AI activity

First seen Nov 10 · 1 event

Henryk Koval reached OpenAI on NW-SALES-03.

Needs review, blocked
Open the demo
One record

One record across browsers, endpoints and agents.

Wherever AI gets used, the record has the same parts: the person, the machine, the browser or agent, the provider, and the processors that provider lists in public.

  1. Alice ChenPerson
  2. nw-alice-mbpDevice
  3. ChromeApplication
    Scout for Browser
  4. ChatGPTProvider
    OpenAI
Listed processorsAmazon Web ServicesGoogle CloudSnowflakeOpenAICloudflareDatabricks

Drawn from the demo organization. Open the demo · What each sensor sees

Privacy

Labels, not content.

No keystrokes. No screenshots. Openwake keeps what kind of data was involved, not what people wrote, unless an owner turns on the Evidence Vault.

What Openwake collects
  • No keystrokes, no screenshots
  • The kind of data is worked out on the machine, and only the label is sent by default
  • Content is kept only if an owner turns on the Evidence Vault
The public record

151 vendors, every fact quoted.

Whether each vendor says it trains on customer data, who is underneath it, and when its terms last moved, quoted from its own documents. Where a document could not be read, the record says unknown, never zero.

See your first AI activity today.

Install Openwake on up to 5 endpoints free and review your first AI activity. After that, Team is $2.50 an endpoint a month. No card required.