Most people think of advanced AI as something that writes, codes, or answers questions. But frontier models like Claude, GPT-5.6, Gemini, and Grok are starting to take on much harder work, including the math behind encryption. These systems can examine how security is built, test possible attacks, compare enormous numbers of ideas, and keep searching longer than any human research team realistically could. Anthropic recently said one of its models found a weakness in the HAWK cryptographic scheme after about 60 hours, even though the system had already gone through roughly two years of expert review. That does not mean AI can suddenly break every bank account, password, or encrypted message. It does suggest that hidden flaws may be discovered faster, putting more pressure on companies to update security before old protections fall behind. The challenge is that stronger encryption still has to run inside payment cards, phones, routers, vehicles, factories, and other systems that may...