Anthropic Warns AI Could Replicate

AnthropicAnthropicAmerican artificial intelligence corporation warned that small AI models could be months away from copying themselves across internet-connected devices without human help. A self-spreading AI worm could be hard to stop because it might run on many devices with no central shutoff. The company said it has not shown a fully autonomous rogue replicator. Its warning is based on agents completing more steps needed to find devices, deliver a model and run it there. Anthropic said smart-home and small-business devices may be early targets because they are online, weakly secured and gaining local AI chips. It urged organizations to harden systems and plan for systems that persist. The tests leave how replication would work in practice unresolved.