The new politics of frontier AI

The United States government has taken a significant step toward easing restrictions on advanced artificial intelligence models, allowing companies like Anthropic to resume global access to their most powerful tools. On June 30, 2026, Anthropic announced that it would begin restoring access to its Fable 5 and Mythos 5 models following a decision by the Department of Commerce to lift export controls previously imposed on these technologies. The move marks a reversal of policies enacted earlier in the month, which had restricted the availability of these models due to national security concerns.

The lifting of restrictions came after a period of heightened scrutiny from the Trump administration, which had been monitoring the potential risks posed by advanced AI systems. Over the past few weeks, the administration had raised alarms about the possibility of these models being misused for cyber threats or other malicious activities. Initially, the government had ordered Anthropic to suspend access to Fable 5 and Mythos 5 on June 12, citing vulnerabilities in the safeguards designed to prevent unauthorized use. This decision was made after the company failed to meet certain security criteria deemed necessary by federal officials.

In response to these developments, Anthropic had initially allowed a select group of American cybersecurity firms to access Mythos 5, demonstrating its willingness to cooperate with regulatory demands. However, the broader public access to the models was halted until the government’s stance changed. On June 26, Secretary of Commerce Howard Lutnick sent a letter to Anthropic indicating that the Trump administration had withdrawn its prior restrictions, signaling a shift in policy. The letter noted that the administration was content with the steps taken by Anthropic to mitigate risks associated with the models.

This change in approach aligns with a broader executive order issued by the Trump administration on June 2, which aimed to establish a collaborative framework between the federal government and private sector entities for the responsible development and deployment of advanced AI systems. The order emphasized the need for joint efforts to ensure that the benefits of AI innovation are realized while safeguarding against potential harms.

The implications of this policy shift extend beyond just Anthropic. Competitors such as OpenAI have also adapted to similar constraints, restricting the release of their latest models, GPT-5.6, to a limited number of approved partners. Despite these limitations, OpenAI's CEO, Sam Altman, expressed dissatisfaction with the process, highlighting the desire for a more streamlined and transparent approach to model releases.

As the landscape of AI governance continues to evolve, the actions of the Trump administration underscore the growing intersection between technological advancement and national security considerations. The situation reflects a complex interplay of corporate interests, governmental oversight, and the ethical responsibilities inherent in developing highly capable AI systems. With the ongoing dialogue between regulators and industry leaders, the future of AI development remains a topic of intense debate and scrutiny.