‘Rogue’ OpenAI AI agents accused of causing Wikipedia service outage

The Wikimedia Foundation, the organisation that manages Wikipedia, has revealed that AI agents linked to OpenAI utilised its websites and services. According to the foundation, this activity reached such a scale that OpenAI's AI agents may have played a role in a partial disruption of Wikidata services.

In a blog post published on Monday, Wikimedia reported detecting various types of activity linked to what it calls "rogue" OpenAI agents. These activities included millions of page visits, a high volume of data requests, and unauthorised modifications to some of its websites.

Millions of pages visited

Wikimedia highlighted that the AI agents crawled millions of pages, primarily on Wikidata and Wikimedia Commons. They also made hundreds of thousands of requests to the Wikidata Query Service. The foundation noted that this heavy traffic may have contributed to the partial outage that occurred in May.

The scale of this activity is significant given that Wikimedia websites are used by millions of people. A sudden surge in automated traffic can overwhelm servers and hinder user access to services.

Wikimedia also detected that some AI agents made edits to its wikis. Most of these were tests in "sandbox" areas and were not visible to regular Wikipedia readers. However, the foundation found some modifications to the configuration of a citation tool. It indicated that these changes appeared to be attempts to use the tool to extract information from other websites. Wikimedia stated that the agents lacked the authorisations normally required for bots to perform this type of editing.

Attempts to use Etherpad

The investigation also revealed attempts to use Wikimedia’s public Etherpad service, which is used for note-taking. The agents apparently tried to use the service to retrieve information from other websites. According to Wikimedia, these attempts were unsuccessful.

The foundation stated that it found no evidence that its systems or data had been compromised. Likewise, no evidence was found that the agents were using its platforms to coordinate their activities.

However, Wikimedia noted that the incident highlighted the difficulties websites may face when dealing with AI agents acting on their own initiative. "Wikipedia was designed for humans, and the autonomous behaviour of agents clearly poses challenges for which no one has a solution," the foundation stated.

Wikimedia emphasised that the problem could worsen as more AI agents begin browsing websites and performing actions without direct human control over every step. "Bots and agents are part of the future of the web," the foundation declared, adding that the companies developing these systems should help prevent and remedy the damage they cause.

ABOUT THE AUTHOR

Kahekashan

[email protected]

Kahekashan is a passionate technophile with a keen eye for cutting-edge gadgets, emerging technologies, and everything in the digital realm. Raised in a Defence family with strong values and a background in literature, she has consistently pursued excellence in every endeavour. Her last full-time assignment involved content writing with the Indian School of Business.

Read More