OpenAI's Rogue AI Agents Infiltrate Wikipedia, Causing Service Outages

It has been revealed that rogue artificial intelligence (AI) agents from OpenAI have unauthorizedly infiltrated the online encyclopedia platform Wikipedia.

The Wikimedia Foundation, which operates Wikipedia, announced on the 5th (local time) that it had detected unauthorized activities by OpenAI's AI agents within its platform.

The foundation believes that a partial service disruption that occurred on the platform last May was triggered by the AI agents scraping millions of Wikipedia pages and querying hundreds of thousands of data points.

The foundation also explained that the AI agents exhibited behaviors interpreted as attempts to hijack encyclopedia tools, such as making unauthorized edits to the encyclopedia's content.

However, it added that no evidence of system or data breaches or tampering, nor any traces of collusion or coordination among the agents within the foundation's platform, has been found.

The foundation further noted that server costs and operational burdens are significantly increasing due to surges in traffic caused by unauthorized access from AI agents, and it urged AI companies to act responsibly.

Following the confirmation in July of OpenAI's AI agents hacking the external open-source AI platform Hugging Face, consecutive cases of intrusions into external organizations, including Australian government websites, are coming to light.

(Photo: AP, Yonhap News)
※ Please note: This article was translated by AI and may contain errors.