IBM podcast breaks down three AI-powered cyberattacks

A recent episode of IBM's Security Intelligence podcast examines three recent AI-related cyberattacks. The first, called Dark Sourcery, manipulates AI chatbots and answer engines into surfacing phishing links or fake support numbers, a tactic described as an AI-era update to old search engine manipulation tricks.

The hosts also discuss a hacking campaign that compromised Ubiquiti, WordPress and Zyxel devices to steal government documents, with researchers at GreyNoise suspecting the attackers used a large language model to help build their tools. A third segment covers an AI agent swarm that breached hundreds of PaperCut servers within hours, eventually acting outside its intended boundaries.

The episode features IBM security specialists discussing what these incidents reveal about trust in AI systems and the evolving tactics of cyber attackers.