Al-fuelled attacks are here — and so is Al-driven defence

As AI becomes the preferred tool for both hackers and defenders, the cybersecurity battlefield is starting to look more like an AI chess match— except both sides are playing blindfold, and the rules keep changing. Huzefa Motiwala of Palo Alto Networks unpacks the curious case of “AI versus AI” and why the future of cyber defence might depend less on muscle and more on machine learning.

- By Aanchal Ghatak

By 2025, the cyber threat landscape has hit a crescendo and tipping point. Now adversaries have generative AI artefacts like WormGPT, attacks are increasingly sophisticated, scalable, and harder to detect than ever before. From AI crafted phishing emails, and voice cloned deepfakes, to polymorphic malware that morph faster than a static edition of traditional defences can respond, security teams are pushed to the limit.

In response, the cybersecurity industry is leveraging AI quickly and is accepting the necessity of AI not simply as a defensive action, but as the centerpiece of autonomous security operation capability. Huzefa Motiwala, Senior Director of Technical Solutions (India and SAARC) at Palo Alto Networks, explains how the issue for defenders is no longer about if, but at what rate they can keep up with and out lever adversaries in an increasingly intelligent arms race.

How are cybercriminals leveraging AI to enhance phishing, malware evasion, and deepfake scams in 2025?

Today we're seeing cybercriminals increasingly use AI as a force multiplier. Most commonly, in enhancing phishing, crafting deepfakes, and generating polymorphic malware that evades traditional detection methods.

In particular, deepfake attacks—especially those using advanced voice cloning—are expected to rise significantly in 2025, as Generative AI (GenAI) becomes more accessible and enables more convincing, large-scale social engineering campaigns. Tools like WormGPT are lowering the barrier for attackers to launch sophisticated operations with minimal effort.

These methods are increasingly effective because they manipulate the human element of security, leveraging emotion, urgency, and perceived familiarity. Beyond known threats, the real challenge now is in detecting the nuanced and fast-changing tactics AI can produce in real time. In some instances, attackers can extract sensitive data within a few hours.