Trump administration lifts AI export controls on Anthropic but tightens pre-screening

The Trump administration has agreed to fully lift export restrictions on Anthropic's cutting-edge AI models, 'Mythos 5' and 'Phaedra 5.' This decision comes just 18 days after the Commerce Department announced sweeping export controls that sent ripples through the global AI industry. Anthropic confirmed on June 30 that it had received notification from the Commerce Department to lift the controls on Phaedra 5 and Mythos 5.

Consequently, the company plans to begin restoring access to these models starting July 1, after integrating enhanced safety features. Phaedra 5, a publicly accessible version of the high-performance AI model Mythos, had raised concerns about potential misuse by hackers for cyberattacks due to bypassable safety mechanisms. The Commerce Department's letter outlined requirements for Anthropic to proactively detect and respond to security risks, develop safety procedures and operational standards in conjunction with the government, and report any malicious uses of the models, such as cyberattacks.

Commerce Secretary Howard Luttwick stated on X that the department worked closely with Anthropic over the past two weeks to analyze Phaedra 5 and approve its release, thereby aligning government positions and strengthening U.S. AI leadership. Anthropic expressed gratitude to users for their patience and everyone involved in facilitating the model's redeployment.

“

Over the past two weeks, we have worked closely with Anthropic to analyze Phaedra 5 and approve its release. This has helped align positions within the government and strengthen America's leadership in artificial intelligence.

Previously, the White House imposed export controls on Mythos 5 and Phaedra 5 on June 12, citing national security. This restricted access not only internationally but also for foreign nationals within the U.S. A partial relaxation followed on June 26, allowing access to Mythos 5 only for "certain trusted partners" approved by the government, paving the way for negotiations with Anthropic. Competitor OpenAI is also limiting initial access to its latest model, 'GPT-5.6,' to select customers at the U.S. government's request.

The Trump administration, established in January last year, had largely maintained a stance of minimal regulation on AI development, emphasizing competition with China in technological prowess. However, the policy direction shifted with the signing of an executive order last month, requiring AI companies to voluntarily undergo government pre-review of new models up to 30 days before their public release. The Wall Street Journal reported that the White House now intends to involve cybersecurity and national security experts more actively in assessing the safety and potential threats of AI models. While the National Institute of Standards and Technology's Center for AI Safety and Innovation (CAISI) has been conducting pre-assessments, national security agencies are expected to directly oversee this process going forward.

“

We thank our users for their patience and everyone who worked to redeploy the model.