An employee can link a personal AI assistant to their work email and files with a single approval, and the company administrator usually does not see the link.
More and more people are using a personal AI assistant—ChatGPT, Copilot, or an assistant installed on a Mac—and, to make things go faster, they give it access to their work accounts. The process involves a simple “allow access” authorization; once granted, the assistant reads emails from Gmail or Microsoft 365 on the employee’s behalf.
For a small business, the problem isn’t an attack, but a lack of visibility. The assistant uses the employee’s account, so in the logs, everything looks like the employee’s normal activity. No one “hacks” anything—the access was granted willingly. The assistant sees everything the employee sees: email, documents, and the calendar.
The second risk arises when an employee leaves the company: deactivating the account does not automatically revoke the approvals and keys granted to an application, so access may remain active after the employee has left.
The good news is that security doesn't cost anything—it's just a matter of a few settings and habits.
What you need to do:
- In Microsoft 365 or Google Workspace administration, you must approve before an employee can link a third-party app to company accounts—this is the setting that determines who can connect an AI tool.
- Check the apps that have already been approved and revoke access to any you don’t recognize.
- Add “revoke access to apps and AI tools” to the checklist when an employee leaves, not just “deactivate the account.”
Sources: The Hacker News | https://thehackernews.com/2026/10/bring-your-own-agent-your-employees.html
This article was generated with AI assistance.