Pwn2Own Winner Chompie Warns AI Could Upend $1.3 Million Human Hacking Contest
Updated
Updated · BBC.com · May 27
Pwn2Own Winner Chompie Warns AI Could Upend $1.3 Million Human Hacking Contest
6 articles · Updated · BBC.com · May 27
$1.3 million was awarded at Pwn2Own Berlin after hackers uncovered 47 new exploit methods, but top individual winner Valentina "Chompie" Palmiotti said she entered thinking it might be her last realistic shot.
Mythos is driving that fear: Anthropic says the model found 1,600 vulnerabilities across hundreds of programs, and Chompie said tools at that level could wipe out the easier bugs that sustain competitions and bug bounties.
Chompie still uses AI as an accelerator—alongside all-night "zombie hacker mode" work that helped her win $20,000 on an Nvidia-linked target and $50,000 on a Linux system—but said only elite researchers may keep pace.
Orange Tsai, whose team won $375,000 in Berlin, took a less bleak view, calling AI a powerful assistant that raises the bar while leaving room for human creativity and intuition.
That shift could still favor defenders: Chompie said AI may make offensive hacking harder overall if the strongest tools reach trusted security researchers before criminals.
As AI automates hacking, are elite cyber experts witnessing the end of their profession?
A new AI can hack systems in minutes. Is Wall Street prepared for the chaos that could follow?
Anthropic’s AI finds flaws everywhere but in itself. Can we trust it to secure global finance?
Pwn2Own Berlin 2026: $600,000 in Prizes Highlight AI Vulnerabilities and the Dual Role of Artificial Intelligence in Cybersecurity
Overview
Pwn2Own Berlin 2026, held from May 14-16, marked a pivotal moment in vulnerability research by ushering in a new era focused on artificial intelligence systems. The event underscored the growing importance of securing AI technologies, as researchers uncovered emerging vulnerabilities in advanced AI products like the NVIDIA Megatron Bridge, which was successfully exploited multiple times. At the same time, the competition demonstrated that traditional software and operating systems still face persistent security challenges. This shift highlights how both AI and conventional platforms require robust defenses as the cybersecurity landscape evolves.