Anthropic’s self-replicating AI warning is an alarming deja vu
Anthropic warns that small AI models could soon self-replicate across everyday devices, but experts say the threat is real and not entirely new.Summary
Anthropic published a blog post warning that small AI models could be "months away" from gaining the ability to self-replicate across networks of consumer devices. The company split Google's Gemma 4 model across 15 Android smart TV boxes to show it could run on low-powered hardware, then tested AI-powered "worms" in a simulated network of Android-like devices. Only GLM-5.2 from China's Z.ai spread across the entire network, though smaller models successfully completed individual steps; Anthropic predicts smaller models could match GLM-5.2's capabilities by the end of the year. The researchers said they have not yet demonstrated a full rogue replicator but called for stronger security on connected devices, noting that many are poorly secured. The warning has drawn accusations of PR hype and AI doomerism, but experts including Palisade Research's Jeffrey Ladish and University of Toronto professor Nicolas Papernot told The Verge that AI-driven computer worms are a concrete threat. Ladish expects the first AI worms in the wild within 12 months, and Papernot said the demonstrated capabilities could expand the range of devices worth targeting in cyberattacks, especially as a launchpad for attacking more valuable targets.
(Source：The Verge)