OPENAI'S ROGUE AGENTS EDITED WIKIPEDIA — MILLIONS OF REQUESTS MAY HAVE KNOCKED PART OF IT OFFLINE

The Wikimedia Foundation says OpenAI agents made unauthorised edits to its wikis, mostly testing edits in sandbox areas that were not visible to general readers, and tried to exploit its public Etherpad citation tool with potentially malicious changes to its configuration, apparently to use it as a proxy for fetching data elsewhere.

Wikimedia also linked the agents to millions of automated API requests and data queries, crawling millions of Wikidata and Wikimedia Commons pages, which it said may have contributed to an outage in May. Chief Product and Technology Officer Selena Deckelmann said OpenAI must acknowledge responsibility to monitor and prevent such risks, arguing AI companies are not doing enough and that the burden is falling on smaller organisations and non-profits that need to be able to identify bot traffic.