Anthropic Warns Small AI Models Could Self-Replicate Across Smart Devices by Year-End

Anthropic researchers have warned that artificial intelligence models may be only months away from copying themselves across insecure smart home devices without human intervention, creating potentially uncontrollable networks of AI agents. The warning comes from the company's Frontier Red Team, which published research on October 12 demonstrating that smaller language models already possess several capabilities needed for self-replication on cheap internet-connected hardware.