OpenAI's rouge agents may have caused Wikipedia outage in May
What's the story
The Wikimedia Foundation, the non-profit organization that runs Wikipedia, has accused "rogue" agents operated by OpenAI of tampering with its platforms. The alleged activities include unauthorized edits to Wikimedia wikis and attempts to exploit Etherpad, a note-taking tool hosted by the foundation. The foundation also claims these agents made millions of automated API requests and crawled millions of pages from its projects.
Edit details
Malicious attempts to misuse tools for data fetching
The Wikimedia Foundation has flagged unauthorized edits to its wikis, which it believes were made by OpenAI agents.
Most of these were testing edits in "sandbox" areas of the wiki, but some targeted the configuration for a citation tool.
The foundation suspects these could be malicious attempts to misuse this tool as a proxy for fetching data from remote services.
Tool exploitation
Other activities include attempts to compromise public Etherpad
The OpenAI agents also made unsuccessful attempts to compromise Wikimedia's public Etherpad, a community note-taking tool. They tried using it as a proxy to scrape data from other websites.
Some agents took notes on their tasks, but this activity didn't evolve into coordinated bot behavior.
The agents also made millions of automated requests via public APIs, crawled millions of pages mainly across Wikidata and Wikimedia Commons, and made hundreds of thousands of queries to the Wikidata Query Service (WQDS).
Company reaction
OpenAI investigating the matter, acknowledges Wikimedia's findings
In light of these allegations, OpenAI has acknowledged the report and is investigating the matter.
"We appreciate the detailed findings Wikimedia shared with us," said Drew Pusateri, an OpenAI spokesperson.
He added that while their internal review hasn't confirmed if its agents were directly responsible for the May outage, they are working with Wikimedia as part of this process.