Conversation review
Check generative AI usage compliance every day
What members send to the AI is checked against your policies once a day. Findings are collected in a list, and admins and managers review and resolve them. It does not stop messages from being sent; it is how you find out afterward.
Open Samurai AppRequest materials
- How the work was done
An AI usage policy, but no way to know if it is followed#
No idea what was sent
Even if customer data or credentials were sent to generative AI, there is no way to find out later.
A policy that stays on paper
You can write an AI usage policy, but there is no way to compare it with how AI is used day to day.
Too many AI usage logs to read
Having someone read AI usage logs one by one stops working as more people use AI.
Check what was sent to the AI against your policies, daily#
Once an admin turns review on, messages sent from then on are checked every day. Findings appear on the Findings tab, where admins and managers resolve each one as a confirmed violation or a false positive.
- Covers what members send to the AI. The AI's replies are not reviewed
- While review is on, a marker appears by the conversation box and shows the details when pressed
- Conversations from before it was turned on are not reviewed
Five ready-made policies, plus your own#
Five policies come ready: customer personal data, sensitive ID numbers, credentials, unpublished financial information, and discriminatory or harassing language. Each can be turned on or off, and you can add your own policies in plain text.
- All settings sit on the Policies and review tab
- Findings keep no copy of the message. Confirmed findings are deleted along with the conversation
Separate what is announced from who can open it#
When something new is found in a one-to-one conversation, the member is told only the policy name, and admins and managers only the count. Only admins and managers can open the excerpt, the reason, and the conversation itself.
- Members get an in-conversation notice, a push, and an email. Admins and managers get a push and an email
- Other members see list items as Hidden because you do not have permission
Frequently asked questions#
Can inappropriate messages be blocked before sending?
No. Conversation review is after-the-fact detection and never stops sending. To check files and text before they go out, use PII detection.
Where can I see AI usage logs?
Review findings are listed on the Findings tab of the Compliance settings screen. Actions are recorded on the Audit log tab, kept for two years in a tamper-evident form.
Can admins read members' conversations?
Only if an admin turns on Member conversations, and then only one-to-one conversations of people in lower roles. It is off by default, and members can check whether theirs may be read. Opening the list or a conversation is recorded in the audit log.
Are conversations from before setup reviewed?
No. Only messages sent after review is turned on are checked. While it is on, a marker appears by the conversation box, so people can see that review is active.
Everyday work comes together, connects, and becomes the company's strength.