AI models enable autonomous cyberattacks and advanced risk management

Companies are facing a new era of autonomous cyberattacks driven by advanced artificial intelligence. A KPMG Cyber Intelligence report highlights the capabilities of Claude Mythos, an AI model developed by Anthropic. Currently accessible to a limited group of companies protecting critical infrastructure—including AWS, Apple, Google, and Microsoft—the model can independently analyze code, identify vulnerabilities, and transform them into functional attack tools.

Claude Mythos has demonstrated the ability to perform tasks that previously required weeks of human labor, such as identifying a 27-year-old OpenBSD bug and a long-standing FFmpeg vulnerability. In testing, it identified 181 working security flaws in Firefox. Unlike previous models that assisted specialists with sub-tasks, Mythos can conduct full attack simulations autonomously.

In a related trend regarding corporate resilience, an EY study involving 1,200 risk management leaders suggests that AI is becoming a vital tool for navigating global crises. While many companies use AI primarily for efficiency, leaders who integrate it into strategic decision-making are better positioned to identify early warning signs of interconnected global risks, such as supply chain disruptions or geopolitical tensions.