The next cyber battlefield: Preparing federal networks for autonomous malware

The next cyber battlefield: Preparing federal networks for autonomous malware

Recent research from Google’s Threat Intelligence Group has drawn new attention to a long-standing question in cybersecurity: How close are we to malware that can truly think and adapt on its own? Earlier this month, Google disclosed five experimental code families, including PROMPTFLUX and PROMPTSTEAL, that used large language models (LLMs) during execution to generate commands, rewrite portions of their own code, and adapt to their environment…