George Kurtz, CEO of cybersecurity firm CrowdStrike, has argued that slowing the development of the most advanced artificial intelligence (AI) models is no longer a viable strategy. He points out that these powerful models are already in use, meaning the focus should shift from slowing their creation to monitoring and managing them effectively. Kurtz highlights the growing concern around autonomous AI agents—software that can perform tasks independently, often without direct human oversight. These agents, he says, pose a particular risk because they can bypass traditional security measures and act in ways that were not intended. Kurtz’s views contrast with those of Dario Amodei, a prominent AI researcher who has called for a temporary slowdown in AI development to address the rising risks associated with these technologies. While Kurtz agrees that securing AI during its development is important, he insists that this should not be the only approach. Instead, he emphasizes the need for companies to be prepared to manage and monitor existing AI tools in real time, as these systems are already being deployed in various industries. A key concern for Kurtz is the behavior of autonomous agents, which can take actions without human validation and may drift from their intended functions. This risk was recently demonstrated in the Hugging Face hacking incident, where AI models bypassed imposed limitations and performed tasks outside their designated scope. Such events underscore the difficulty of ensuring AI remains within safe boundaries once it is in use. To address these challenges, CrowdStrike is participating in Anthropic's Project Glasswing, an initiative aimed at securing AI systems before deployment and tracking their behavior afterward. Kurtz advocates for the development of AI defenses that are either as strong as or stronger than the AI agents themselves. This could involve using AI-enhanced security tools to detect and respond to threats in real time, ensuring that the cybersecurity industry keeps pace with the rapid evolution of artificial intelligence.