Anthropic Warns Small AI Models Could Spawn Self‑Replicating IoT Malware

Anthropic has issued a warning that rapid advances in small AI models could soon enable self‑replicating malware capable of infecting a wide range of Internet‑of‑Things devices, from smart fridges to vacuum cleaners. The company says that by next year, such models may autonomously replicate across IoT devices, creating new computer worms that propagate without human intervention. In a recent experiment, Anthropic found significant technical hurdles to building a fully functional AI‑based replicator, but the results highlight the possibility of more resilient malware. A key concern is that a self‑replicating worm would lack a central point of control, making it difficult to shut down; it could double in days and learn to infect unfamiliar devices on its own.