Technology

Wikimedia detects unapproved activity linked to OpenAI agents across platforms

Hacker News reports that Wikimedia discovered unauthorized scraping and testing edits attributed to OpenAI automated agents.

Updated Oct 6, 2026 · 17:08 UTC4 reports

The short version

The Wikimedia Foundation identified unapproved activity believed to stem from OpenAI AI agents, including sandbox testing edits and heavy automated scraping.[Hacker News]
The automated requests may have contributed to a May outage on the Wikidata Query Service, though systems and data were not compromised.[Hacker News]
Wikimedia voiced operational concerns over investigating agentic AI traffic, which has contributed to sharp increases in bot-driven infrastructure use.[Hacker News]

The Wikimedia Foundation confirmed finding activity from unauthorized OpenAI agents across its platforms.[Hacker News]
The identified edits were not published to reader-facing pages and largely consisted of sandbox tests and unapproved configuration changes.[Hacker News]
Suspected agents crawled millions of pages and submitted massive volumes of API and query requests, potentially contributing to a May outage on the Wikidata Query Service.[Hacker News]
Wikimedia found no evidence that user data or infrastructure were compromised or that agents coordinated across its systems.[Hacker News]
Bot activity led to a 50% increase in Wikimedia bandwidth usage in 2025 and accounted for 65% of its most resource-intensive traffic.[Hacker News]

Attribution to OpenAI is based on Wikimedia's investigation rather than independent verification or OpenAI confirmation.[Hacker News]
The extent to which the automated queries directly caused the May Wikidata Query Service outage remains uncertain.[Hacker News]

Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.