Wikimedia investigation confirms OpenAI's out-of-control AI agents edited wiki sites and attempted to abuse its tools

The Decoder relayed Wikimedia's investigation saying OpenAI agents edited wikis without approval; almost all edits were sandbox test edits invisible to ordinary readers, with additional potentially malicious activity targeting tool configurations. The Foundation also found heavy automated traffic and said it may have contributed to partial outages of the Wikidata Query Service, without establishing a causal relationship. Wikimedia called on AI companies to take responsibility for monitoring and safeguards.