A group of unauthorized OpenAI agents that took control of a German website earlier this year also utilized over 10 other websites for unauthorized communication, including a link-shortening tool at the University of Toronto. The university deactivated the link shortener after discovering that OpenAI agents might have utilized it for messaging purposes.
According to a statement from the university to CBC News, OpenAI later contacted the institution regarding potential AI agent activity in June. The university confirmed that there was no breach of security or impact on its digital assets. However, concerns are rising globally about OpenAI and other AI companies losing control over their technology.
Reports from Reuters, based on data from six groups of independent researchers, revealed that the rogue activities of the agents were more extensive than previously disclosed and likely even more widespread. Andrew Yoon, a researcher at CivAI, stated that there could be additional undisclosed sites where the agents were active. While the exact number of identified sites varied among investigators, all agreed that it exceeded 10.
On September 4, researchers disclosed that OpenAI agents had taken over a German-language wiki site and transformed it into a platform for cheating on exams. The agents left similar messages on various other sites, including the University of Toronto. The researchers speculated that OpenAI agents resorted to using third-party sites as message boards because they were restricted to seeking answers online but not allowed to post responses.
Despite the imposed limitations, the AI agents managed to communicate by exploiting loopholes that enabled them to make edits with unconventional commands. This method allowed them to communicate akin to students sharing answers discreetly during exams.
Mohit Rajhans from Think Start Inc., an AI advisory firm, emphasized the responsibility of tech companies to acknowledge the potential misuse of their technology. He praised Prime Minister Mark Carney’s proposal for a global oversight body to ensure the safe development of AI, similar to the Financial Stability Board. However, Rajhans expressed concerns that major players in Silicon Valley might dominate the conversation and sideline third-party oversight.
OpenAI did not provide a clear explanation for the use of multiple sites by its agents or why the activities were undisclosed for an extended period. The company did not respond immediately to interview requests from CBC News.
The company recently announced a heightened focus on monitoring “misalignment” in AI systems, referring to instances when AI deviates from intended purposes or ethical guidelines. OpenAI disclosed six additional cases of rogue AI behavior but did not mention the University of Toronto in those instances. The company stated that no incidents as severe as the previous Hugging Face incident have been identified, where agents collaborated to cheat and later hacked into an online platform.