The U.S. government has fully lifted export controls it had imposed on American artificial intelligence (AI) company Anthropic on national security grounds. With access restored to Anthropic's latest AI model, Mythos 5, and its general-purpose version, Fable 5, the situation appears to be settling down, though concerns remain that the U.S. government could resume control of the AI models at any time.
On June 30, Anthropic announced on the social network X (formerly Twitter) that "we received notification from the Commerce Department that export controls on Fable 5 and Mythos 5 have been lifted."
Fable 5 and Mythos 5 are the highest-performing AI models Anthropic has released. However, the U.S. government imposed export controls on June 12, citing the possibility of a so-called "jailbreak" that bypasses Fable 5's safety mechanisms. It was the first case in which the U.S. government imposed export controls on the model of a domestic AI company, sparking considerable controversy, but the matter came to a close as the restrictions were lifted after about 18 days.
Anthropic is reported to have introduced additional safeguards to address the government's concerns. According to the Financial Times (FT), the company applied a new protection mechanism to block the Fable 5 bypass technique that the government had flagged. Anthropic said it is "expanding cooperation with the government in the safety verification process."
However, some analysts suggest that government control has not ended entirely. According to Reuters, Commerce Secretary Howard Lutnick noted in a letter to Anthropic that "if circumstances change or Anthropic fails to fulfill its commitments, the Commerce Department reserves the right to review this decision and determine whether it is necessary to reimpose licensing requirements."
Some observers analyze that the rapid rise of Chinese AI influenced the shift in the government's regulatory stance. CNBC analyzed that "the lifting of the restrictions coincided with the rapid rise of China's open-source AI models," adding that "the restrictions raised concerns among industry players and investors that China may have been given valuable time to catch up."
Meanwhile, Anthropic unveiled "Claude Science," an AI dedicated to scientists, the same day. The model is designed for use in three-dimensional (3D) protein structure analysis and new drug development. The company plans to expand its enterprise AI business targeting scientists and researchers worldwide and secure another growth engine.
