Microsoft Finds Phishing Operators Repurposing an AI Jailbreak Trick to Slip Millions of Emails Past Filters

Microsoft security researchers say spammers have adapted a Unicode-based trick first developed to jailbreak AI chatbots, hiding invisible characters inside financial keywords to push millions of phishing emails past automated filters. The technique, known as ASCII smuggling, was detailed in a Microsoft Security Blog post published September 3, 2026 by researchers Noam Kochavi and Sarah Wolstencroft, who traced a finance-themed phishing operation whose volume jumped from a low-thousands daily baseline to nearly 1.3 million messages overnight before peaking above 2.37 million.