OMM Redaksiyası
Small artificial intelligence models could soon be able to copy themselves across internet-connected devices without human control or access to data centres, according to research by “Anthropic”.
The company’s Frontier Red Team study, published on Sunday, examined whether AI models could find devices, transfer their files, install the software needed to operate and continue spreading from each newly infected machine.
Models with fewer than 30 billion parameters can already complete many individual stages of the process, the researchers said, although most still struggle to combine them reliably.
GLM-5.2 spread to eight devices in a controlled test
In a simplified experiment, “GLM-5.2” spread to all eight available test devices in four of five trials, and reached seven devices in the remaining trial.
“Anthropic” said the experiment simplified the hacking stage by using devices with unsecured connections. The results do not show that an AI model can currently escape and reproduce autonomously in real-world conditions.
“We don’t yet have a full rogue replicator,” the researchers said, while warning that AI agents were becoming better at completing the necessary steps and passing useful information to later copies of themselves.
The study focused on inexpensive internet-connected devices, which often have limited security protections. Potential targets include smart televisions, refrigerators, ovens, vacuum cleaners, security cameras and videoconferencing equipment.
Researchers demonstrated that a 26-billion-parameter model could run across 15 inexpensive “Android” TV boxes, each costing about $55. A smaller model successfully installed the software needed to operate on eight of the nine types of hardware tested.
Unlike conventional computer worms, a self-replicating AI system could potentially adapt to unfamiliar devices, diagnose technical problems and pass successful strategies to future copies, the researchers said.
Such a system would initially spread slowly, possibly taking one or two days to create each new copy. However, it could be difficult to eliminate because it would have no central server that could be switched off.
“Anthropic” estimated that, if smaller models continue improving at their historical pace, they could reach the capability demonstrated by “GLM-5.2” by the end of 2026. The company cautioned that this was a rough projection based on general capability trends.
The researchers recommended monitoring unusual network traffic, isolating infected devices and requiring manufacturers to strengthen default security settings. They said early, slow-moving replicators could be a sign of faster systems capable of breaching more secure computers.
“We should aim to harden our systems to make it difficult for such replicators to spread,” they said, “but also prepare to deal with a world where they are here to stay.”
