Anthropic has called out Smart for the first time, accusing the GLM-5.2 distilled Claude and OpenAI models

According to Sentinel Beating monitoring, Anthropic's National Security Policy Lead Tarun Chhabra publicly mentioned SmartSpectrum for the first time at the Aspen Security Forum. He accused GLM-5.2 of distilling Claude and OpenAI models.

Distillation is the process of training one model with the responses of a strong model. The technique itself is common, but the controversy lies in whether it circumvents terms of service by scraping competitors' outputs on a large scale.

Chhabra stated that the top AI labs in China are almost exclusively distilling American models. Anthropic is currently banning millions of relevant accounts weekly. The leading-edge American models are currently ahead of China by approximately 6 to 9 months. He believes that without this type of distillation, the gap could have been 12 to 18 months.

Prior to this, Anthropic had previously called out DeepSeek, DarkSide of the Moon, MiniMax, and Alibaba.