The Wikimedia Foundation called on AI companies on October 5 to make their agents identifiable to website owners and help prevent unwanted activity. Chief Product and Technology Officer Selena Deckelmann published an operator report on unapproved bot activity across Wikimedia services.
The foundation reported wiki edits without required bot approval and unsuccessful attempts to use its Etherpad note-taking service as a fetching proxy. Almost all edits were sandbox tests; Wikimedia said they did not appear on pages visible to general readers. It believes OpenAI agents were responsible. That attribution has not been independently confirmed.
Wikimedia said it found no evidence of compromised systems or data, or coordination among agents. Whether heavy downloading contributed to an outage in May remained open. Some published links to citation-tool configuration changes now lead to deleted revisions, preventing complete technical reconstruction.
OpenAI spokesperson Drew Pusateri told Reuters and The Verge that the company was reviewing the findings with Wikimedia and would share further information. According to The Verge, OpenAI had not verified whether its bots contributed to the May outage. The updated article explicitly notes the added company statement.
Wikimedia’s request targets AI-system operators: website owners need to recognize and prevent unwanted agent activity. The foundation wants to stop cleanup after such activity becoming routine work for the people maintaining the open web.