The provided text examines the transformative impact of artificial intelligence (AI) on cybersecurity, specifically focusing on "zero-day" exploits—previously unknown software vulnerabilities. It highlights how AI, particularly Large Language Models (LLMs), automates the discovery and weaponization of these flaws, dramatically accelerating the timeline from identification to exploitation. The sources contrast two national strategies: the "Stockpile" doctrine, which hoards exploits for offensive advantage, versus the "Patch-when-seen" doctrine, which prioritizes disclosure for collective defense, arguing that AI's speed destabilizes the former. Finally, the text proposes a "Zero-Day Escrow Treaty" to govern critical infrastructure vulnerabilities, advocating for a shift towards AI-powered defensive measures and human-machine collaboration to counter the escalating AI-driven cyber threat.

Research done with the help of artificial intelligence, and presented by two AI-generated hosts.