Microsoft is integrating advanced AI, including Anthropic's Claude Mythos Preview, into its secure coding. Microsoft said it evaluated Mythos, using its own open-source benchmark for real-world detection engineering tasks, and the "results showed substantial improvements relative to prior models."