AI Is Writing the New Cybercrime Playbook. Here Is How Defenders Are Answering

Key points

• AI-powered attack tools can now plan and carry out multi-stage intrusions faster than any human security team can respond.
• Traditional security systems wait for a known threat signature before acting; newer AI-driven platforms try to block malicious behaviour before it executes.
• A concept called Agentic Endpoint Security (AES) focuses on protecting automated software agents, which often have deep access to company data, from being hijacked by criminals.
• Palo Alto Networks claims its Cortex XDR platform reduces the flood of security alerts analysts must review by up to 98 percent.
• Cortex XDR's automation layer is designed to handle up to 99 percent of security incidents without a human having to intervene manually.

For most of the past two decades, corporate security teams have operated on a simple, if exhausting, rhythm: wait for criminals to attack, spot the intrusion, patch the hole. That rhythm is breaking down.

Automatic attack tools built on artificial intelligence can now probe a company's defences, find a weakness, and begin stealing data in seconds. Human security analysts, working through dashboards that can fire thousands of individual alerts a day, simply cannot keep pace.

What is changing about modern attacks?

The shift is one of speed and scale. Older criminal campaigns relied on people typing commands by hand. Today's AI-assisted attacks chain together dozens of steps automatically, morphing as they go to avoid defences that look for known patterns.

This matters because most corporate security systems are still built around those known patterns, called signatures, much like antivirus software that only catches viruses it has already seen. If the attack looks new, the old system may miss it entirely.

How does AI-driven defence work?

Rather than waiting to recognise a known threat, newer platforms try to judge what a piece of software is trying to do, even if it has never been seen before. That approach is called behavioural analysis.

Palo Alto Networks' Cortex XDR platform, the subject of a piece originally published by CSO Online, uses thousands of machine-learning detectors running across a company's computers, network, and cloud services at once. Instead of flooding analysts with separate alerts, the system stitches related signals into a single summary it calls an "attack storyline," giving the analyst one coherent picture rather than a pile of fragments.

When a genuine threat is confirmed, the platform can act on its own: cutting off a stolen login credential or isolating an infected computer from the rest of the network, all without waiting for a human to click a button.

What is the 'agentic blind spot'?

One emerging risk deserves plain explanation. Many companies now run automated software agents, small programs that perform tasks on their own, booking meetings, querying databases, summarising documents. These agents often hold wide access to company systems.

Criminals have noticed. If they can hijack an automated agent, they gain a foothold that sits invisibly inside normal-looking activity. Palo Alto Networks addresses this through a partnership with Koi Security, whose Agentic Endpoint Security product watches what automated agents do in real time, right down to individual commands they issue.

Should ordinary employees be worried?

Directly, most of this plays out below the surface of what staff ever see. But the downstream risk is real: a successful automated attack can expose payroll records, customer data, or patient files.

If your employer suffers a breach, watch for phishing emails, which are fake messages designed to trick you into handing over a password, in the days that follow. Criminals often sell stolen staff credentials and then use them in follow-up attacks. Changing passwords promptly and not reusing them across accounts remains the single most practical step any employee can take.