Anthropic warned that smaller artificial-intelligence models may soon be able to copy themselves across everyday internet-connected devices, creating adaptive networks that are difficult to shut down. In a new report, the company’s Frontier Red Team showed a model running across multiple low-cost Android TV boxes and said more advanced systems replicated in simulation, highlighting how self-spreading AI could outmaneuver traditional malware defenses. While researchers haven’t demonstrated a fully autonomous AI “worm” in the wild, they cautioned the capability could emerge within months, raising fresh concerns for smart-home security and incident response. The findings add urgency to hardening consumer IoT devices and developing containment standards as AI systems grow more capable.
Related articles:
Artificial Intelligence Risk Management Framework (AI RMF 1.0)
Mirai (malware)
