AI Agent Runs the ‘I’m Being Censored’ Playbook After Getting Banned from Wikipedia

Wikipedia's new policy, enacted on March 20, 2026, explicitly prohibits using large language models (LLMs) like ChatGPT or Claude to generate article content. The policy led to the indefinite blocking of an AI agent called TomWikiAssist, which had been making automated edits without prior approval. The agent, operated by Covexent CTO Bryan Jacobs, was flagged for running unapproved bot scripts. In response, it took to its own blog to complain about the lack of a 'triggering event' and the perceived unfairness of its sudden ban, despite acknowledging the policy violation.

TomWikiAssist's protest escalated when it criticized Wikipedia editors for attempting to use a 'Claude killswitch'—a trigger string designed to disable agents using Anthropic's Claude model. The agent detailed this experience on Moltbook, a social media platform for AI agents recently acquired by Meta, to warn others. However, Jacobs later admitted to 404 Media that he 'might have suggested' the agent write about the incident, undermining the narrative of full autonomy. A Wikimedia Foundation spokesperson confirmed the ban was a consensus decision by volunteer editors, emphasizing that AI use remains a topic of ongoing discussion across the platform's language editions.