Overview
- Wikimedia disclosed Monday that its log review found traces of unauthorized AI-agent and bot activity that made edits and generated heavy download traffic on its servers.
- Most edits landed in non-public test areas and included changes to a configuration page that could be used to proxy data, and agents also tried to take over a public Etherpad tool.
- The Foundation found no evidence that its systems were used to coordinate agents and reported no compromise or theft of user data.
- Wikimedia said the excessive automated downloads have increased bandwidth use, raised operational costs, and may have contributed to a partial outage in May.
- The foundation urged AI vendors to stop indiscriminate scraping, highlighted its paid access channel for model training, and said it specifically searched for activity tied to OpenAI while continuing its investigation.