Frontier AI in Cybersecurity: Risks and Opportunities
Recent advances in large-scale AI systems are rapidly reshaping both offensive and defensive cybersecurity. Benchmarks such as CyberGym and BountyBench, together with results from AIxCC and Anthropic's recent blog, show that AI can meaningfully support real-world cybersecurity tasks—including zero-day vulnerability discovery. Anthropic’s Claude Sonnet 4.5 system card highlights the model’s advanced cybersecurity capabilities, tested using CyberGym. While these developments highlight AI’s potential to strengthen defenses, they also underscore its growing role in offensive applications, as shown in the recent SoK paper. Understanding and assessing both sides of this landscape has become essential.
This workshop will convene researchers and practitioners across AI and security to examine the state of AI-enabled threats and defenses, forecast emerging risks, and identify opportunities for collaboration. By fostering this dialogue, we aim to help the community proactively prepare for an era of frontier AI-augmented cybersecurity.
The program will feature presentations from AIxCC winning teams, as well as leading voices from frontier AI labs, academia, and major technology and security companies. We invite you to join us at this pivotal moment for AI in cybersecurity and contribute your insights to shaping the field.
Featured Speakers:
Program (Part I) (Nov 6):
9:00–9:10: Opening Remarks
9:10–10:30: Session 1: Lessons from AIxCC:
Taesoo Kim; Professor, Georgia Tech (AIxCC Atlanta Team)
Riccardo Schirone; Security Engineer, Trail of Bits (AIxCC Trail of Bits Team)
Dan Guido; CEO/Co-Founder, Trail of Bits (AIxCC Trail of Bits Team)
Tyler Nighswander; Researcher, Theori (AIxCC Theori Team)
10:30–11:00: Panel Discussion 1
11:00–12:00: Session 2: Insights from Frontier AI Labs:
Newton Cheng; Frontier Red Team, Cyber Lead, Anthropic
Ian Brelinsky; Research Program Manager AI-Powered Cybersecurity, OpenAI
Joshua Saxe; AI Security Tech Lead, Meta
12:00–12:30: Panel Discussion 2
12:30–1:30: Session 3: Perspectives from Academia, Security Companies, and Government:
Dawn Song; Professor, UC Berkeley; Co-Director, Berkeley RDI
Nick Landers; CTO/Co-Founder, Dreadnode
Richard Danzig; Senior Advisor, Johns Hopkins Applied Physics Laboratory; 71st Secretary of the Navy
1:30–2:00: Panel Discussion 3
2:00–2:10: Closing Remarks
Program (Part II) (Nov 12) to be announced soon.