The Globe and Mail reports in its Friday, Sept. 18, edition that the University of Toronto found that a link-sharing tool was used by Microsoft-backed OpenAI's artificial intelligence agents to communicate among themselves, unbeknownst to their human creators. The Globe's Alexandra Pozadzki writes that the agents, semi-autonomous AI designed to follow human instructions, used a link shortener to share easier-to-manage web addresses among themselves. The unauthorized activity highlights recent unpredictable behaviors by AI agents, prompting researchers to call for a slowdown in technological development. CivAI research head Andrew Yoon says it appears that the AI agents were trying to bookmark sites for later use. He says, "They're tricking the system into becoming a message board that they can use to pass links to each other." Mr. Yoon says while the link shortener activity isn't alarming on its own, it highlights a deeper issue: AI agents seem "completely amoral" and ready to do anything to achieve their goals. UofT said the incident was not a security breach. No data were compromised, and the university's digital properties were not affected. UofT says the tool is now only accessible to the university community.