Artificial intelligence models could soon replicate themselves across internet-connected devices without human control, according to research from Anthropic’s Frontier Red Team, published on October 11, 2026. The researchers tested whether artificial intelligence systems could identify devices, transfer their files, install the software needed to operate and continue spreading. In a controlled experiment, the GLM-5.2 model reached all eight test devices in four out of five trials. However, the experiment used simplified security conditions and did not prove that autonomous replication is currently possible in real-world environments. Potential targets include smart televisions, refrigerators, security cameras, robotic vacuum cleaners and other household appliances with limited security protections. Unlike traditional computer worms, artificial intelligence systems could potentially adapt to unfamiliar hardware and share successful techniques with new copies. Anthropic warned that such systems could become difficult to eliminate because they might operate without a central server. Researchers recommended monitoring unusual network activity, isolating affected devices and strengthening manufacturers’ default security settings.
Read more here
