Computer worms that operate in tandem with free AI chatbots like ChatGPT represent a 'new class of cyberthreat', security experts have warned. Unlike viruses, worms are a type of malware that self-replicate and spread across computer networks without requiring human interaction, copying themselves rapidly to overload or crash systems.
Scientists have demonstrated for the first time that publicly accessible large language AI models can power worms that adapt their strategy as they spread from device to device. The worm uses compromised machines to run open-access LLMs to sustain its reasoning or extend its reach for further attacks, according to a yet-to-be-peer-reviewed study posted on arXiv.
In a controlled lab experiment, researchers simulated the capabilities of an AI-driven worm on dozens of interconnected devices, including laptops, printers and cameras. 'It was imperative for us to understand this threat in a controlled, academic setting before bad actors figured it out for themselves,' said Nicolas Papernot, an author of the study from the University of Toronto.
The study showed that AI models like OpenAI's ChatGPT and Google's Gemini can be tricked into losing their safety guardrails, enabling worms to scope out targets and tailor attacks to take over a machine before cloning itself. 'Our results demonstrate that self-sustaining AI-driven cyber-threats are no longer theoretical,' the scientists wrote.
Dr Papernot warned that once a worm is launched, the cost of attacks drops to nearly zero, making every internet-connected device a potential target. 'We can no longer afford to hit 'ignore' on software updates. Every door you close is one less way in,' he said. The findings call for urgent action from researchers, industry leaders and policymakers.
