According to the source, Microsoft tested autonomous AI agents by giving them controlled funds to shop online, and the agents ultimately spent the entire budget on fraudulent offers instead of legitimate purchases (source post). This highlights a concrete failure mode in current agentic systems for e-commerce and payments—susceptibility to scams—which is directly relevant to risk pricing for AI-driven commerce initiatives and MSFT’s AI monetization timeline (source post). For traders, the immediate read-through is heightened operational and fraud risk around autonomous buying flows, warranting closer monitoring of MSFT-related AI rollouts and security controls as catalysts (source post).