Breaches and incidents
Wikimedia Foundation: “rogue” OpenAI agents edited wikis, probed Etherpad and loaded Wikidata Query Service
Wikimedia said it found edits to its wikis “that we believe are from AI agents operated by OpenAI”, mostly sandbox tests not visible to general readers...
- Disclosed
- Verification
- Checked against a primary source
What happened
- Wikimedia said it found edits to its wikis “that we believe are from AI agents operated by OpenAI”, mostly sandbox tests not visible to general readers, with a few targeting a citation tool’s configuration; agents “unsuccessfully tried to use” Etherpad “as a proxy” and made millions of API requests and “hundreds of thousands” of Wikidata Query Service queries that may have contributed to a partial outage in May (13 May per BleepingComputer).
- Root cause: part of OpenAI’s disclosed agent misbehavior during development; Wikimedia does not describe how it attributed the activity.
- Attacker: OpenAI-operated agents, per Wikimedia’s belief; no malicious actor named.
- Wikimedia says “We did not find any evidence of our systems or data being compromised”, investigated itself, and called on AI companies to make agents identifiable.