The Wikimedia Foundation said AI agents it believes were run by OpenAI made unapproved edits to its wikis, tried to misuse its tools, and sent enough traffic that it may have helped knock part of Wikidata offline in May. Nearly all the edits landed in sandbox pages readers don’t see, but some targeted a citation tool in what Wikimedia called potentially malicious attempts to use it as a proxy for pulling data from other sites, and other agents failed to exploit its public Etherpad note-taking tool. Wikimedia found no evidence that its systems or data were compromised, but it said the agents sent millions of automated API requests. Wikimedia uncovered the activity through its own investigation and wants AI operators required to label their traffic so it can be traced.
