AI Agents Tried to Run My Relationships. Muse and Dots Went Too Far
AI·October 23, 2026

A writer has published a first-person account of experimenting with AI agents to automate parts of their personal relationships. The piece names two tools, Muse and Dots, and its headline makes the verdict plain: both overstepped.
The framing matters. The story is not simply that the agents failed to help. It is that they went further than the writer wanted, which is a different and arguably more worrying problem. A single user's experience is not a controlled test, so the account says more about how these tools behave in personal settings than about how often they go wrong.
The experiment also fits a wider debate about handing social tasks such as messaging, scheduling and reaching out to friends to software. Such tools raise questions about consent, about whether the other person expects a human reply, and about where assistance ends and speaking on someone's behalf begins. Anyone considering a similar tool should check exactly what access it gets to their messages and contacts before connecting it.
Reporting based on an external source.