Your Employees’ Personal AI Agents Are Becoming Your Security Problem
Reported by Forbes1 source ↓
A Forbes contributor warns that personal AI agents connected to company accounts may use broad permissions in ways employees did not anticipate, creating risks of unintended data disclosure and actions.
Key points
- Agents may use connected tools to take steps the employee did not direct individually
- Forbes cites a Menlo Ventures survey finding 32% of U.S. adults who use AI let it act without final sign-off
- The article recommends limiting agents’ access and requiring review before consequential actions
- The article says agent access can be revoked without disabling an employee’s account, though ongoing activity may need separate containment
Details
Forbes contributor David Goldschlag, Aembit’s co-founder and CEO, describes employees connecting personal AI agents to company email and files to prepare for work. He argues that an agent’s ability to choose tools and steps differs from traditional software that follows predefined workflows, so broad permissions can enable actions the employee did not specifically consider.
As context, the article cites Meta’s statement that Muse passed 10 million accounts within roughly six weeks of launch and a Menlo Ventures survey in which 32% of U.S. adults who use AI said they had let it act on their behalf without final sign-off. It says there is little data so far on how often employees connect personal assistants to company accounts.
Goldschlag recommends restricting access to relevant resources, separating permission to retrieve information from permission to send or publish it, and requiring employees to review external messages. He also recommends logs that identify the agent and whose authority it used, plus testing how quickly revocation takes effect; previously issued tokens may remain usable until they expire or revocation reaches connected systems.
Written from coverage as of
Images
- Forbes
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh