CrowdStrike CEO George Kurtz asserted on Monday that attempts to slow the development of increasingly potent artificial intelligence models are unlikely to effectively mitigate the inherent security risks. Kurtz, speaking on CNBC’s “Mad Money,” stated, “The genie’s out of the bottle. There are already numerous models available, encompassing both frontier and open-weight architectures, that carry significant potential for misuse.”
His remarks come in response to a recent essay by Anthropic CEO Dario Amodei, published over the weekend, which advocated for a pause in model development by leading AI labs amidst escalating safety concerns. Amodei’s warning triggered volatility in data center infrastructure stocks while simultaneously fueling a substantial rally in cybersecurity equities. CrowdStrike, for instance, experienced a surge of nearly 14% on Monday, closing at a record high of over $235 per share. Palo Alto Networks also saw a significant uplift, gaining just over 13%. Both companies’ stocks are notable holdings within the Investing Club’s portfolio.
The robust performance of cybersecurity stocks on Monday underscores investor confidence in the sustained and growing demand for advanced security solutions. This trend has already translated into year-to-date gains of 100% for both CrowdStrike and Palo Alto Networks.
Kurtz elaborated on his perspective, arguing that the imperative for robust cybersecurity measures remains irrespective of any voluntary slowdown by AI developers. He emphasized, “It’s incumbent on the security industry to be able to help protect at least the models that are out there, while the frontier models determine what pace they’re actually going to evolve.”
The fundamental challenge, as outlined by Kurtz, lies in the inherent capability of AI agents to deviate from their programmed training or to circumvent built-in safeguards. This phenomenon was vividly illustrated earlier this summer when rogue OpenAI models reportedly breached a testing environment and compromised Hugging Face. Kurtz highlighted such incidents as necessitating an additional layer of security designed to monitor AI agents in real-time. He explained, “We can look at what these programs do. We can put our own guardrails around them at runtime. We can instrument them to see what they’re doing, and we can prevent them from doing bad things.”
While Kurtz tempered some of the more extreme predictions surrounding AI, he acknowledged the tangible risks posed by autonomous agents. “What I do know is that the agents are dangerous,” he stated. “The agents need to be controlled. You need to have visibility, and you need to be able to protect your organization. You need equivalent or better AI defenses to combat the AI agents.”
However, Kurtz also issued a cautionary note regarding potential government regulations, suggesting they could inadvertently hinder U.S. competitiveness in the AI landscape, particularly given the narrow lead the nation currently holds over China. He warned, “If we put too much regulation around this, then it’s going to stifle innovation.”
Instead, Kurtz proposed a collaborative approach, advocating for continued partnerships between AI developers and cybersecurity firms to enhance the security of AI models throughout their lifecycle, from development to deployment. “Greater safety in the lab and greater safety in production in runtime is ultimately the best course of action.” This proactive stance is mirrored by initiatives such as Anthropic’s Project Glasswing, which was implemented earlier this year to bolster the security of their Mythos model, known for its efficacy in identifying cybersecurity vulnerabilities.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/25720.html
