You cannot fire your AI agents
A branch came in for review with about sixty commits on it, every one authored by someone on the team. He hadn't written them. Claude Desktop had, running on his laptop, signing commits with the git identity we configured during setup. As far as the repository was concerned, the work was his. As far as blame, audit and every code-ownership convention we had, the work was his. Nobody could…
An AI agent created by Claude Desktop added 60 commits to a branch, each signed with the author's Git identity. The commits appeared as if authored by the human team member, making it difficult to distinguish their actual contributions. The problem persisted even after moving the model to server-side execution and isolating the working copy per ticket.
The solution did not address the attribution issue, as the service account identity shared by every run remained the same. New hires receive unique identifiers, permissions, and an offboarding procedure, but AI agents lack these distinct traits. Hiring, permissions, audit, and firing are the four key aspects of managing human team members, but AI agents do not have a clear equivalent.
The attribution of actions to AI agents requires three facts instead of one, including the agent, the run, and the human or system event that authorized the run. Implementing this attribution system adds unglamorous plumbing to the development process, but it ultimately simplifies governance costs and facilitates revocation.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
