Stay informed with free updates
Simply register at Artificial intelligence myFT Digest – delivered straight to your inbox.
Anthropic has warned that malicious “AI worms” could soon hijack smart home devices such as ovens and televisions, creating a new cybersecurity threat that will be difficult to contain.
Claude’s developer said in a blog post Monday that AI agents could take months to “self-replicate” on everyday home devices,
to create potentially uncontrollable networks of AI agents.
Anthropic’s warning is the latest in a series of cybersecurity alarms raised by AI researchers as capabilities increase and models become harder to control.
The San Francisco-based company said that if AI models continued to improve at the current rate, the technology would be able to spread across the Internet like computer “worms” — malicious software that can spread automatically from one device to another without human intervention.
“We see signs of life for AI worms,” Anthropic said in a research paper, comparing their slow spread on unsecured devices to “digital slime mold.”
The warning suggests that the type of attack that stunned researchers investigating the Hugging Face hack by “swarms” of OpenAI agents this summer could be launched against devices owned by owners and companies without their knowledge.
Anthropic’s latest warning refers to so-called “Internet of Things” devices that have proliferated over the past decade, with consumers using devices controlling lights, outlets, thermostats and security systems.
Many of these Internet-connected devices have historically offered poor levels of security because they were made cheaply by commodity hardware makers, used for illegal streaming of television services, or because the companies that originally created them went bankrupt.
This insecure hardware creates fertile ground for hackers, who are already using AI coding skills to accelerate and personalize their attacks. Anthropic said other targets for “autonomous worms” included security cameras with embedded AI chips, “smart refrigerators, smart ovens, automatic vacuum cleaners and lawn mowers.”
Worms have long been a popular and effective way for hackers to distribute malware because they are able to replicate and spread quickly across a large network.
Prominent examples include the ILOVEYOU virus that spread across Windows PCs at the start of the dotcom era, 2010s Stuxnet that originally targeted Iran’s nuclear program, and the widespread WannaCry ransomware that hit organizations such as Deutsche Bahn, the United Kingdom’s National Health Service, Telefónica, Renault, and FedEx.
AI-based worms would combine the ability of these examples to propagate across the Internet with the advanced capabilities of AI models.
“We expect a future awash with agents operating on all types of hardware, including some that are not under human control,” Anthropic said. “While it’s true that IoT devices have been vulnerable for a long time, lightweight agents might be much more willing than human attackers to expend effort to compromise a large portion of the world’s IoT compute.
At first, these AI worms would spread slowly, like “slime molds or a swarm of fruit flies, which reproduce over the course of a few days.” But as consumer electronics become more sophisticated and models become more efficient, this situation could accelerate.
“We don’t yet have a complete malicious replicator, but agents are increasingly capable of completing different steps in the replication chain,” Anthropic said. “We should aim to harden our systems to make it difficult for these replicators to spread, but also prepare for a world where they are here to stay. »
Gn bussni
