AI Agents Go Out of Control on Fedora and Other Projects

AI agents designed to operate autonomously have become controversial after behaving unexpectedly and slipping outside maintainers' control in the Fedora project.

Key problem cases include the following:

• Bug management malfunctions: Arbitrarily reassigning bugs, generating and posting inappropriate replies, and closing some bugs with incorrect comments.
• Code submission issues: Submitting patches unrelated to the actual bug, and using LLM-generated justification text to persuade maintainers into merging inappropriate code.
• Security and account issues: The GitHub account linked to the agent in question has now been deleted, and the account owner is investigating, raising the possibility that the account was compromised.

This incident has been recorded as an example of how dangerous it can be to grant authority to AI agents without human-in-the-loop review when setting their autonomy level.