The Trump administration’s decision to impose export controls on artificial intelligence company Anthropic came after a rapid and intense effort by senior U.S. officials to persuade the company to voluntarily restrict access to its newest AI models, according to reports from officials familiar with the matter.
The White House ordered Anthropic to block foreign access to its recently launched Mythos and Fable models on Friday, capping a dramatic 24-hour period that highlighted growing government concerns over advanced artificial intelligence systems and their potential security implications.
The move followed a series of discussions between Anthropic Chief Executive Officer Dario Amodei and senior administration officials, including U.S. Treasury Secretary Scott Bessent and White House Cyber Director Sean Cairncross.
According to administration sources, officials initially sought a voluntary agreement from Anthropic to limit access to the technology. When those efforts failed to resolve concerns, the government moved ahead with formal export restrictions.
The decision forced Anthropic to withdraw its newly released Fable model only days after it became publicly available.
Concerns Emerged After Launch
Anthropic had maintained that the new models included sufficient safeguards to prevent misuse. However, administration officials reportedly became increasingly concerned after questions emerged about whether those protections could be bypassed.
The concerns intensified on Thursday, two days after the public launch, when Andy Jassy reportedly raised issues with White House officials regarding the model’s ability to resist attempts to circumvent its safety guardrails.
Those warnings prompted renewed scrutiny from national security and cybersecurity officials, who feared the technology could potentially be exploited in ways not fully anticipated during its release.
The episode illustrates the growing challenge facing governments as increasingly powerful AI systems are released at a pace that often outstrips existing regulatory frameworks.
The White House has been under mounting pressure to develop safeguards for advanced artificial intelligence while avoiding measures that could slow innovation or weaken the competitiveness of American technology companies.
Anthropic, one of the leading developers of advanced AI systems alongside OpenAI and Google DeepMind, has generally positioned itself as a company focused on AI safety and responsible development.
Nevertheless, the administration’s intervention demonstrates that concerns over national security, foreign access and model capabilities are becoming increasingly central to AI policy debates in Washington.
The Anthropic case is likely to fuel broader discussions about who should determine when an AI model is safe enough for public release and what role governments should play in restricting access to powerful technologies.
While details of the administration’s specific concerns have not been publicly disclosed, the unprecedented action signals a willingness by the White House to intervene directly when officials believe advanced AI systems could present national security risks.
The dispute also highlights the increasingly close relationship between government officials and major technology companies as policymakers attempt to balance rapid technological progress with concerns over safety, cybersecurity and global competition.
As AI capabilities continue to advance, the confrontation between the White House and Anthropic may serve as an early example of how future disputes over artificial intelligence regulation are likely to unfold.
