Security ·
Wikimedia: AI agents made unauthorized edits and flooded servers
The Wikimedia Foundation published a report on agents it believes were operated by OpenAI, which made unapproved edits on its platforms and sent millions of requests.
Most edits were tests that ordinary readers never saw. Some, however, touched the configuration of a citation tool, which the Foundation considers potentially malicious. The agents also tried, unsuccessfully, to compromise a public Etherpad. The traffic may have contributed to a partial Wikidata Query Service outage in May.
The Foundation found no evidence that data was compromised. According to Reuters, OpenAI has not confirmed involvement but is cooperating to verify the activity.
What it means for companies
The case shows why an agent needs clear limits: what it may do, where it may go and how it identifies itself. Without rules and a record of each step, even a well-meant agent can cause damage.
Sources: BleepingComputer, Cybernews