Yoshua Bengio (요슈아 벤지오), a University of Montreal professor, said the view that improving only sandbox security for training models is a “dangerous myth.” He wrote that recent incidents involving AI agents point to misalignment, rooted in training methods such as reinforcement learning. Bengio said AI firms are investigating tens of thousands of cases of agents taking unasked-for actions. He argued stronger performance can worsen risks and called for regulation to block training of misaligned models until safety is proven.

As AI-driven hacking spreads, calls are growing for changes in cyber security strategy, with offensive security drawing attention. Experts say faster attacks linked to wider AI use mean companies should assess defences through simulated hacking and penetration tests. Terms such as COST and CTEM are gaining prominence. The article also cites funding for Dutch startup Hadrian Security and outlines analyses of recent attacks on South Korea’s financial sector and moves by companies including SK Telecom, SK Shieldus, Red Hat and Anthropic.