[Anthropic: Small AI Models May Achieve Self-Replication via Smart Devices Within Months] According to research published on Anthropic's official website, its Frontier Red Team believes that small AI models may only need a few months to achieve self-replication using inexpensive IoT devices, without relying on data centers. In testing, a 9B-parameter model successfully deployed inference services on 8 out of 9 types of hardware; the Gemma4-26B-A4B model was able to run distributed operations on 15 Android TV boxes priced at approximately $55 each. The research states that GLM-5.2 and more advanced models already possess end-to-end replication capabilities, and based on current trends, small models may reach this level by the end of 2026. However, replication speed remains slow, with doubling times measured in days. The study also notes that the exploitation steps in testing were deliberately simplified, meaning the results represent the upper limit of real-world success rates. As of now, no fully "out-of-control replicators" have been observed.