From human speed to machine speed

Artificial intelligence is changing the cybersecurity landscape, making sophisticated cyberattacks faster, more accessible, and increasingly difficult to defend against. But while the technology is introducing new challenges, many of the underlying security problems remain familiar.

That was a central message of “The AI Cyber Challenge,” an opening keynote presentation and panel discussion at Carnegie Mellon University's 2026 CyLab Partners Conference, where researchers and industry leaders explored how AI is reshaping cybersecurity and what organizations must do to keep pace.

Vyas Sekar, co-director of CyLab's Cyber Autonomy Initiative, opened the session by examining the growing capabilities of AI-powered cyberattacks and separating emerging threats from industry hype.

“What used to be a human time scale problem has become a machine time scale problem,” said Sekar.

While vulnerabilities, multi-stage attacks, and poor cybersecurity practices are nothing new, Sekar emphasized that AI is dramatically accelerating the speed of attacks, lowering the expertise required to carry them out, and exposing weaknesses that organizations could previously overlook.

To address these challenges, Sekar highlighted CyLab's Cyber Autonomy Initiative, which is supported by key industry founding sponsors and aims to establish the scientific foundations for autonomous cybersecurity systems capable of defending against threats at machine speed. He also introduced the Cyber Autonomy Arena, a research platform designed to evaluate autonomous attackers and defenders in realistic environments.

Following the presentation, Lujo Bauer, co-director of the Cyber Autonomy Initiative, moderated a panel featuring Sekar; Sarah Cen, assistant professor at CMU; Harry Krejsa, director of studies at the Carnegie Mellon Institute for Strategy and Technology; and Lea Kissner, chief information security officer and vice president at LinkedIn.

The discussion explored whether AI ultimately favors attackers or defenders, the growing national-security implications of AI-enabled cyberattacks, and whether slowing AI development could provide time to establish stronger safeguards.

“Short term, it's definitely attackers,” said Kissner, noting that organizations are already facing increasing numbers of vulnerabilities and shrinking response times.

However, Kissner expressed optimism that AI could eventually help defenders build more secure software and improve their ability to prevent attacks before they occur.

Panelists also emphasized that technology alone will not solve these challenges. Cen highlighted the importance of responsible AI development, meaningful guardrails, and accountability, while Krejsa stressed the need to make advanced cybersecurity defenses accessible beyond the world's largest and best-resourced organizations.

Despite the challenges ahead, the discussion underscored opportunities for researchers, industry leaders, and policymakers to work together to build more effective, scalable, and responsible cybersecurity defenses.

Watch the full presentation and panel discussion in the video below to hear the experts’ perspectives on the future of AI and cybersecurity, and view the presentation slides for a closer look at the research and ideas discussed during the session.

To learn more about partnering with CMU CyLab and the Cyber Autonomy Initiative, contact Michael Lisanti and the CyLab Partnerships team at partnerships@cylab.cmu.edu.