Skip to content
Northgrove
Menu

Free check

Is your AI agent one email away from leaking data?

Three questions. Thirty seconds. No sign-up.

Private data: Can the agent access anything private, such as customer records, internal documents, an inbox or a CRM?
Untrusted content: Does it read content that someone outside your company can influence, such as incoming emails, tickets, web pages or uploaded files?
Outbound channel: Can it send anything out, such as emails, messages, API calls, links or images, or write to a system other people read?

Answer all three questions to see your result.

Your answers are not stored and are not sent anywhere. The check runs entirely in your browser.

Where this comes from

“The lethal trifecta” is Simon Willison's name for the combination of private data, exposure to untrusted content and a way to send data out. Read his original post. This check covers that one risk. It does not cover jailbreaks, wrong answers or other ways agents fail.