Did China Gain Access to Anthropic's Top-Secret AI?

The White House's decision to impose export restrictions on Anthropic's advanced AI model "Mythos" was reportedly motivated by suspicions of unauthorized access by groups connected to China, according to a new report from U.S. media outlet "Semafor."

According to the report, one of the reasons the White House proceeded with restrictions was the government's grave concern that Mythos may have been unlawfully accessed by groups connected to China. If the Chinese government actually gained access to cutting-edge models such as "Mythos 5" or "Fable 5," it could pose serious risks to U.S. national security.

Of particular concern is the imitation of technology through a technique called "distillation." This involves an AI playing the role of a "student" learning from the outputs of a more advanced model to replicate its behavior, amounting to a form of reverse engineering. If the Chinese government leverages this technique, there is a possibility that it could incorporate the capabilities of cutting-edge AI developed by the United States at great cost and effort into its own models in a short period of time.

Meanwhile, the White House has not officially confirmed the contents of this report. In a statement posted to X (formerly Twitter) by David Sacks, an adviser to President Trump, there was no mention of China, and it only explained the need for export restrictions from another perspective.

Anthropic is one of the leading AI companies representing the United States alongside OpenAI, which developed ChatGPT, and the leakage or misuse of its cutting-edge models could be a critical turning point in the U.S.-China AI hegemony race. If this suspicion proves true, discussions on AI security are expected to accelerate further, and the impact is spreading among industry professionals and policymakers.

At this point, many details remain unclear, and an official explanation from the White House and the parties involved is awaited.