CrowdStrike launches tool to stop out-of-control AI agents
The new feature allows security teams to terminate autonomous software and cut off its network access immediately.
By Kevin Iyer ·
CrowdStrike announced a new capability on Tuesday that enables security teams to shut down artificial intelligence agents that are behaving unexpectedly. The feature, unveiled at the company's Fal.Con Europe conference, gives organizations a direct method to halt specific software processes and prevent them from restarting.
George Kurtz, the chief executive of CrowdStrike, described the new tool as an essential addition to modern security infrastructure. He stated that while many companies are currently building AI agents, they previously lacked a simple mechanism to stop them. "Everyone is building AI agents, but they never had the off button, and now they do," Kurtz told CNBC's Jim Cramer on the show "Mad Money." He added that the feature is a critical component in ensuring these systems operate safely.
The tool is part of the Falcon Guardian platform, which monitors AI agents within an organization. It is available immediately and is designed to address safety concerns as businesses rapidly adopt autonomous systems. The technology allows administrators to terminate misbehaving agents and cut off their network access, giving human operators more control over AI models.
Kurtz noted that the industry has traditionally focused on protecting companies from external threats. He argued that security providers must now also ensure that internal AI systems operate safely. "Enabling agents to work safely is a massive market opportunity for CrowdStrike," he said. He explained that the company uses its existing AI capabilities to prevent external breaches while ensuring agents follow their specific instructions.
The CEO compared the new safeguard to airbags in automobiles. He suggested that even well-designed systems require safety mechanisms. "If they go outside of those guardrails at runtime, we can kill those agents," Kurtz said. This approach aims to mitigate risks associated with the rapid deployment of autonomous software in enterprise environments.
CrowdStrike is seeing increased demand for its security tools as businesses integrate AI. Kurtz pointed to a partnership with Anthropic to protect its Claude Cowork product. He said that enterprise customers requested the integration of CrowdStrike technology before deploying Anthropic's offerings. "Our customers went to Anthropic and said, 'We need CrowdStrike before we can deploy AI,'" he said. He viewed this demand as evidence that customers are reluctant to roll out AI without established security measures.
Stock prices for the company reflect recent market trends. CrowdStrike shares reached a record high of $278.86 on Oct. 6. Since that date, the stock has fallen by approximately 12% as the broader technology sector has experienced volatility in October and November.
