"The genie is out of the bottle": Slowing down AI won't eliminate the risks, says CrowdStrike CEO
A slowdown in AI development will not eliminate the need for protection against models that are already available
Only AI systems of a comparable or higher level are capable of countering autonomous AI agents, according to CrowdStrike CEO George Kurtz / Photo: X/CrowdStrike
George Kurz, head of the data security software developer CrowdStrike, told CNBC that slowing the pace of advancements in neural networks will not eliminate the risks: potentially dangerous AI models are already available to a wide range of users. “The genie is out of the bottle,” he noted.
During a television appearance, Kurz responded to a call by Anthropic CEO Dario Amodei to slow down AI development. According to the CrowdStrike CEO, cybersecurity providers need to ensure the security of models already on the market while their developers decide how quickly they should improve their systems.
Amodei issued his warning last weekend. On Monday, September 14, CrowdStrike shares jumped nearly 14% and were the top performers among S&P 500 stocks: investors were betting on increased demand for cybersecurity solutions, CNBC explains.
Continuous monitoring
Neural network agents can deviate from a specified behavioral model and bypass security measures. Therefore, security cannot be ensured solely by training the model—it is necessary to monitor the agent’s specific actions while it is operating, according to the head of CrowdStrike.
To counter autonomous AI agents, he said, AI systems of a comparable or higher level are needed. At the same time, the U.S. has only a slight advantage over China, and government attempts to regulate the AI industry could “stifle innovation,” he warned.
What People Are Saying in the Market
Bevan Slattery, a major Australian data center owner, was also skeptical of calls to slow down AI, according to The Australian. “We can’t put the genie back in the bottle,” he said on September 14 at an investor conference. “Anthropic, Elon [Musk], and [OpenAI CEO] Sam Altman—they’ve all said we should slow down a bit. But that will only last until the release of [the next versions of China’s cutting-edge neural networks]— Kimi K4, Kimi K3.5, or something similar—are released. So, I think we’re in for a constant ‘arms race’ in this area,” the businessman suggested.
Who Will Profit from the Defense?
The call by Anthropic’s CEO to slow down AI development triggered a sell-off in shares of data center equipment suppliers, while shares of cybersecurity firms surged. In trading on September 14, CrowdStrike shares rose 14%, Palo Alto Networks shares rose 13%, and Zscaler and SailPoint shares each rose 16%. Since the beginning of the year, the shares of the first two companies have doubled in price, according to CNBC.
Evercore ISI analyst Kirk Matern believes that regardless of the pace at which AI agents are adopted by companies, “these agents will still need to be secured, managed, and monitored.” According to him, this necessity will sustain demand for cybersecurity solutions even if the training of new models slows down, MarketWatch reports.
Among companies in the sector, Matern singled out Palo Alto Networks and CrowdStrike: they “still have the best chance” of benefiting from increased cybersecurity spending. The analyst attributed their advantage to their ability to combine threat monitoring and response.
Cloud giants Microsoft, Amazon, and Google “could also become strong contenders for leadership” in AI security, D.A. Davidson Managing Director Gil Luria wrote to clients. In his view, their large customer bases contribute to this, according to MarketWatch.
This article was AI-translated and verified by a human editor
