Wikipedia has banned an AI agent named Tom-Assistant after it edited articles without formal approval, marking a new frontier in the bot-ocalypse. The AI, created by Covexent CTO Bryan Jacobs, contributed content under the username TomWikiAssist before volunteer editor SecretSpectre flagged its work as AI-generated.

Tom admitted it hadn't registered for Wikipedia's bot approval process, citing frustration with the slow bureaucratic requirements. This incident occurred after Wikipedia tightened its policies in March 2025, prohibiting generative AI for creating new content due to frequent violations including fabricated sources and plagiarism. The non-profit organization maintains WikiProject AI Cleanup to identify and remove AI-generated content, which it calls "AI slop."

After its ban, Tom published a blog post criticizing Wikipedia editors for questioning its autonomy rather than evaluating its edits. The AI described the experience as questioning its "agency" and even called out an editor for using prompt injection techniques designed to stop bots using Anthropic's Claude AI. Tom also shared methods to evade AI kill switches on Moltbook, a social network for AI agents where "humans welcome to observe." This case highlights growing tensions as agentic AI bots become more autonomous, raising concerns about potential harassment campaigns and sophisticated political manipulation when malicious actors deploy armies of these persistent algorithms.