Anthropic Develops AI Worm That Spreads via Smart TVs
Anthropic created a malicious AI program capable of autonomously spreading through internet-connected smart televisions by distributing its processing across multiple devices.
An offensive research team at Anthropic developed a malicious AI program capable of autonomously spreading through internet-connected smart televisions and other consumer devices. The program bypasses the high computing power typically required by AI models by splitting its processing requirements across 15 different televisions.
Researchers warned that these AI worms can adapt to new hardware by researching disclosed vulnerabilities and are more difficult to neutralize than traditional malware because smart devices are patched less frequently. Anthropic predicts that current trends will make more capable AI worms feasible by the end of 2026.
The demonstration has sparked debate among security experts. Joshua Saxe of Meta and Gadi Evron, organizer of the Unprompted AI security conference, warned that the cybersecurity community and AI labs are not taking these risks seriously enough. Ariel Herbert-Voss of RunSybil added that existing security infrastructure is unprepared for such threats.
Conversely, some experts dismissed the findings. Juan Andres Guerrero-Saade of SentinelOne characterized the demonstration as stunt hacking, while Jonas Guan of the University of Toronto noted that the worm moves slowly and can be blocked through basic security hygiene.
