Translation result
U.S. cybersecurity firm Palo Alto Networks warned that AI-powered, next-generation hacking attacks are rapidly becoming a reality and urged organizations to take immediate action.
Lee Klarich, Palo Alto’s chief product officer, wrote in a blog post on May 13 that organizations now have roughly three to five months to get ahead of AI-enabled attackers. He warned companies must act quickly to head off an incoming flood of vulnerabilities.
Concerns spiked after Anthropic introduced Mythos, a security-focused AI model. The situation prompted the White House to convene major banks and tech companies for an emergency response meeting.
Google disclosed this week that it blocked an attempted large-scale, AI-driven attack. The company said hackers are already using publicly available AI tools to probe for and exploit software vulnerabilities.
Klarich cautioned that the threat is not limited to future models. “A wide range of publicly available AI tools alone can enable powerful attacks,” he said, urging the industry to develop new defensive technologies.
He highlighted the value of virtual patching — applying temporary protections at the security layer to block exploits before software fixes are deployed. Klarich called it a critical defensive measure for an era of ultra-fast, AI-driven attacks.
Palo Alto Networks plans to roll out related capabilities in the near term, he added.
Anthropic limited access to Mythos to a small group of companies last month. Test partners include Palo Alto, CrowdStrike, Amazon, Apple and JPMorgan Chase.
OpenAI has also entered the AI-security race, unveiling a cybersecurity-focused model called “GPT-5.5-Cyber” and announcing a security project named “Daybreak.”
“Just a few weeks ago, some questioned whether AI models’ security capabilities were overhyped,” Klarich said. “After additional testing, it appears we may have underestimated their performance.”
By Chan Park cpark@aitimes.com
Most Commented