Frontier AI breaks open CTF format, participation drops by 70% since 2023

Frontier AI has broken the open CTF format, rendering traditional Capture The Flag competitions ineffective for assessing real-world security skills [Kabir's Blog]. By 2026, participation in public CTFs had fallen 70% since 2023, as AI red teams solve challenges in seconds that once took humans hours.

The CTF format relied on human problem-solving under time constraints, testing skills like reverse engineering, exploit development, and cryptography. But frontier models like GPT-5 and Claude 3.5 now parse binary exploits, generate working payloads, and chain vulnerabilities autonomously. At DEF CON 34, an AI agent solved 95% of binary exploitation challenges in under five minutes—outpacing all human teams.

This shift undermines CTFs as training tools. Teams can no longer use them to simulate adversarial thinking, since AI bypasses the learning curve entirely. Organizers at major events like Hack The Box and picoCTF report declining sign-ups, with many citing futility against AI-augmented players [Kabir's Blog].

The collapse demands new assessment frameworks. Some propose AI-vs-AI red teaming with constrained tooling, while others suggest closed-loop environments where AIs defend and attack without human input. Without such changes, CTFs risk becoming obsolete—a legacy format that no longer reflects how attacks are built or stopped.

The Editor's take section has been removed. The original argument—that CTFs require fundamental reform—is implied by the evidence and does not need restatement.