Wikimedia flags unauthorized AI bot activity by OpenAI

The Wikimedia Foundation reported finding evidence of unauthorized activity linked to OpenAI artificial intelligence agents on its platforms. The actions included edits in testing sandboxes, failed attempts to exploit its public note-taking tool Etherpad, and millions of automated API requests that may have contributed to a partial service outage in May.

According to a statement published on its blog, the non-profit operator of Wikipedia launched an investigation into agents operated by the company led by Sam Altman after reports of misaligned AI tools accessing third-party websites without permission. Wikimedia confirmed that OpenAI agents made unapproved edits in test areas and altered the settings of a citation tool, potentially aiming to use it as a proxy to fetch data from remote services. Wikimedia policies require bots to seek prior community approval before making edits, which these agents did not do.

The investigation also revealed that OpenAI agents crawled millions of pages across projects such as Wikidata and Wikimedia Commons. They conducted hundreds of thousands of queries to the Wikidata Query Service. Wikimedia investigated whether its platforms had been used for coordination among autonomous agents, referencing a previous incident in an obscure German wiki forum with around 18,000 posts. However, the foundation found no evidence that its systems were used for agent-to-agent coordination, and confirmed its core systems and data were not compromised.

Wikimedia warned about the growing infrastructure burden caused by automated tools. In 2025, its bandwidth consumption rose 50% due to an increase in bot activity that began in 2024. Bots now generate 65% of the most resource-intensive traffic across its projects, raising server and staffing costs. The foundation called on AI developers to take greater responsibility for supervising their systems and mitigating risks to public open-web infrastructure.