Shadow AI
Shadow AI: start with a safe place to use it, not a ban
Ban generative AI and the people who need it start using personal accounts. Samurai App provides a company AI environment, checks on what goes in before it is sent, and records of what happened afterward, all together. IT can widen AI use while seeing how it is used.
Open Samurai AppRequest materials
- Where things are
- How the work was done
Banned, and still in use#
Use on personal accounts
Without an approved environment, work documents get pasted into personally paid generative AI. The company cannot tell what was sent where.
Bans do not last
People who have seen how useful AI is keep using it after a ban. The gap between policy and practice grows.
Input checks left to the sender
You can ask people not to send customer data or API keys, but the only one checking before it goes out is the sender.
Give the company a generative AI environment it can approve#
Samurai App gives each company one tenant and stores its data in a separate database. Input is not used to train AI. Agents act only within the permissions of the person using them.
- Four roles: owner, admin, manager, and member
- Sensitive actions do not run until a person approves them
- Two-factor authentication is required for everyone and cannot be turned off
Check for personal and secret data before it is sent#
Inside Samurai App, files and text bound for the AI are checked in the browser before they are sent. With the browser extension, attachments and pastes on other companies' generative AI sites get the same check. When something is found, the sender decides whether to send.
- Detection uses no model and runs entirely in the browser
- The browser extension supports Chrome and Edge and can be deployed to managed devices through Chrome policy
- Once the extension is connected, sending despite a finding records the site name and count in the audit log
Review what was sent and done, from the records#
With conversation review on, what members send to the AI is checked against your policies every day. Review never stops sending; admins and managers look at the findings. Actions are kept for two years in a tamper-evident audit log.
- Audit logs export as CSV and JSONL, and JSONL can be verified wherever it is taken
- Screening also judges and logs whether imported text carries instructions aimed at the AI
Frequently asked questions#
What is shadow AI?
It is when employees use generative AI for work that the company does not know about. Work documents and customer data can end up in services outside the company's view.
Does banning generative AI prevent data leaks?
A ban alone tends to leave use on personal accounts. Samurai App provides an environment the company can approve, then combines input checks before sending with review and records afterward.
Can sending to other generative AI sites be blocked?
There is no setting that stops sending automatically. With the browser extension for Chrome and Edge, attachments and pastes on other generative AI sites are checked before they are sent. It does not block sending either; the sender decides.
Is the data we enter used to train AI?
No. Training opt-outs are set with the model providers. Skills and memories created from conversations are used only within that workspace.
Everyday work comes together, connects, and becomes the company's strength.