Tipping the Cyber Balance: How AI Benchmarks Could Make Software Safer

"In February 2024, a ransomware attack on Change Healthcare disrupted medical claims processing for nearly half of all U.S. health care transactions. The breach cost UnitedHealth Group over $2.8 billion, exposed the personal data of 190 million Americans, and forced hospitals nationwide to delay patient care. The cause? A remote access portal without multi-factor authentication. As one senator put it: “This hack could have been stopped with cybersecurity 101.”

This attack illustrates a broader pattern. Critical infrastructure depends on complex systems with sprawling attack surfaces—misconfigurations, excessive privileges, inadequate monitoring, and software vulnerabilities—and attackers are exploiting these weaknesses faster than defenders can address them. Artificial intelligence is accelerating this dynamic: The same technology that helps developers build applications faster also enables attackers to find and exploit flaws more quickly..."

Source: RAND

Read more here: https://www.rand.org/pubs/commentary/2026/02/tipping-the-cyber-balance-how-ai-benchmarks-could-make.html