Your Staff's Personal AI Agents Are Probably Already Logged In to Work Systems

Personal AI agents are arriving quickly, and the rollout is already bumpy. According to the article, Amazon now blocks Meta's Muse because it does not identify itself as a bot while browsing, privacy concerns continue to surround another agent called Instinct, and stories of agents going wrong keep piling up. Even so, large tech companies are backing the idea that everyone needs a personal assistant, less than a year after the first release of the open-source agent now called OpenClaw.

The article's central warning is that employees will connect these agents to work systems, with or without approval. Agents that already manage their home schedules will naturally move on to work schedules. The author compares this to Bring-Your-Own-Device, but says it is a bigger challenge: these are not just insecure devices holding company data, but AIs acting on company systems in the employee's name. A personal agent also cannot be enrolled like a laptop, because it runs on the vendor's infrastructure or blends into the employee's own environment.

The article frames this as an identity problem first. The best case is an agent with an identity that has a named owner and limited approval, though even then it may never be offboarded. Worse cases include agents posing as employees and agents with no clear owner. Rather than simply trying to block them, the author recommends enabling safe use with identity as the control point. People typically connect these agents the way they connect any other app, without raising a ticket.