Anthropic warns small AI models may soon spread across devices

anthropic researchers william zhou and daniel freeman tested small language models on off-the-shelf android tv boxes to see if they could spread on their own across poorly secured iot devices. they found that the glm-5.2 model can complete the full replication chain across all eight test nodes. while no current model is both small enough and capable enough to do this in the wild, researchers expect this gap to close soon. anthropic ceo dario amodei warns of potential damages in the hundreds of billions of dollars. this could lead to ai worms that are slow but hard to kill.