Trump drops restrictions on Anthropic’s Mythos and Fable models

In a significant shift, the U.S. government has lifted the requirement for Anthropic to obtain a license before exporting its advanced AI models, Mythos and Fable. This decision restores public access to some of the most sophisticated AI technologies available, which had been previously restricted due to export regulations. Starting July 1, Anthropic plans to reinstate access to these models following a period of negotiations with government officials. The restrictions had been imposed on June 12, classifying the AI models as export-controlled technologies, which complicated their availability to international users. The inability to comply with these rules at scale led the company to halt public access entirely. Commerce Secretary Howard Lutnick announced that Anthropic will now take proactive measures to identify and mitigate security risks associated with its AI models. The company has committed to working closely with the U.S. government on safety protocols and to report any malicious use of the technology. Notably, Anthropic had already been voluntarily pledging to implement many of these measures prior to the restrictions being enacted, raising questions among cybersecurity experts about the necessity of the bans. Critics viewed the export limitations as more of a punitive measure against Anthropic for its leadership's comments regarding potential misuse of AI technology by the government and its political adversaries. Initially, Mythos was selectively released to a limited number of organizations in April to address concerns over its potential to uncover and exploit software vulnerabilities. The public release of Fable followed in June, featuring enhanced security measures. As competition heats up in the global AI market, particularly with emerging Asian firms like Fugu and Tulonfeng introducing comparable models, the U.S. faced increasing pressure to relax its restrictions on Anthropic. Just last week, Lutnick authorized the release of Mythos to select clients vetted by the White House. Additionally, OpenAI's latest models have also been made available to a curated group of organizations, rather than the general public. The inconsistent AI policy framework under the Trump administration has created uncertainty for tech companies regarding future model releases. An executive order issued in June, which aimed to review models prior to their deployment, was met with criticism from prominent analysts, including Dean W. Ball, who recently took a policy role at OpenAI.