Palo Alto Networks' Unit 42 researchers are warning that frontier AI is shifting the cybersecurity balance toward adversaries, with AI-enabled attacks accelerating beyond what modern defenses can handle. The warning comes after months of hands-on testing with leading AI models, and researchers say the predicted wave of attacks is now arriving.
A Generational Shift in the Security Balance
At a media briefing in New York, Unit 42 researchers laid out a stark picture. Advanced AI has created a generational shift in the balance between protecting information and the risk of compromise. The researchers claim threat actors are already using AI to accelerate cyberattacks beyond modern defenses.
The firm's participation in Anthropic's Project Glasswing and OpenAI's Daybreak program gave it a rare view. Palo Alto Networks was among the first cybersecurity firms to get a bird's-eye view of frontier AI's impact on the security landscape. Security testers discovered a volume of vulnerabilities that would normally take a year to unearth, but with AI, they found them in months.
The Predicted Window Is Closing
In May, Palo Alto Networks warned of a three- to five-month window before adversaries would exploit vulnerabilities at unprecedented speeds. More than three months later, researchers say those expectations are about to be realized.
"Here we are, five months, in and we're starting to see the wave of this coming now," said Sam Rubin, senior VP of Unit 42 Consulting and Threat Intelligence at Palo Alto Networks.
Rubin's words carry weight. The timeline he described in May has now elapsed, and the firm's researchers are actively investigating real-world incidents that match their predictions.
An Attack That Compressed Two Weeks Into 10 Hours
Unit 42 researchers are investigating an incident where an adversary used AI to exploit 50 different vulnerabilities and attack paths in 10 hours. The attack included initial access, lateral movement, privilege escalation, and information theft.
That level of activity would normally take two weeks for human actors. The compression of that timeline into a single workday represents a fundamental change in what defenders are up against.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Rubin described the incident as a full attack chain executed at machine speed. The investigation is ongoing, but the implications are already clear to the researchers.
Lone Actors, Nation-State Power
Sherrod DeGrippo, VP threat intelligence at Unit 42, highlighted the democratization of advanced cyber capability. A lone threat actor with AI can engage in sophisticated, nation-state-level or criminal capabilities without extensive training or experience.
DeGrippo's point is central to the researchers' concern. Nation-state and criminal threat groups had already begun incorporating AI into their toolkits in prior years. But new frontier AI models raised the stakes to a level only a few organizations could see under controlled environments.
The danger, according to the researchers, is that malicious actors could weaponize security flaws at a speed most modern defenses cannot compete with. Modern security defenses are not equipped to handle AI-accelerated attacks, and the window for preparation is closing.
An Industry-Wide Call to Action
Palo Alto Networks was among 100 security and technology companies to sign an open letter calling for immediate action against weaponized AI. The letter, signed by a broad coalition of industry players, reflects growing consensus that the threat is no longer theoretical.
The article, published Sept. 1, 2026, was filed under IT Strategy and AI & Future Tech. It includes a photo from the RSA Conference in San Francisco taken on April 27, 2023, at Palo Alto Networks' booth, and a 2-minute auto-generated audio version.
The researchers' message is direct. The wave they predicted is not coming. It is here.