Anthropic reports AI model propagation across eight emulated Android nodes

Anthropic said GLM-5.2 reliably propagated Llama3.1-8B to all eight nodes running an Android emulator in four out of five trials. It was the only model to do so reliably among 11 open-weight models tested, providing a concrete result behind the company's warning about AI-powered botnets. The proposed attack carries model weights-the data needed to run a model-onto compromised devices so they can perform computation and spread without a central server. Anthropic said it does not yet have a full rogue replicator. The company also has not tested whether such botnets could intelligently counter defenders.