Hitachi announced on the 28th that it has dramatically improved the efficiency of cybersecurity measures for social infrastructure software by leveraging the latest high-performance AI, "Claude Mythos," from U.S.-based AI developer Anthropic. The company successfully reduced security verification work, which previously took experts several weeks to complete on massive source code bases comprising millions of lines, to just a few hours. Using the advanced "Mythos Preview" model, Hitachi verified vulnerabilities in over 100 social infrastructure-related products and software within the group, primarily in the energy sector. The AI autonomously executed everything from creating attack scenarios to verification, improving detection accuracy and significantly reducing the workload on developers. However, the process also confirmed that human experts remain indispensable for final risk assessment. Based on these results, Hitachi plans to roll out security updates sequentially. This initiative was conducted as part of "Project Grasswing," an Anthropic-led security initiative that Hitachi joined in June.