The Trump administration recently implemented an export ban on Anthropic’s most advanced AI models, citing national security concerns related to a jailbreak (a method to bypass internal safety guardrails). This action forced the company to pull access to its Mythos and Fable 5 models earlier this month. The government directive, which initially gave Anthropic just 90 minutes to comply, originated after a reported disagreement with the Pentagon over modifications to guardrails for military use, leading the Department of Defense to designate Anthropic a supply chain risk. Anthropic maintains that the identified vulnerabilities are minor and comparable to those found in other models, a position supported by several independent AI safety researchers who have reviewed the government’s findings.

The incident highlights the lack of a transparent or consistent regulatory framework for AI in the United States. While the Trump administration released a national policy framework in March advocating for sector-specific regulation rather than a single rule-making body, critics describe the current enforcement approach as ad hoc and opaque. An executive order requiring companies to voluntarily share advanced models for government vetting was delayed due to concerns that it might impede innovation. Meanwhile, the administration is navigating a fragmented legal landscape, with states like California and Florida pursuing their own AI safety laws and litigation.

Anthropic, an AI firm nearing a public offering with a valuation approaching $1 trillion, remains central to these tensions. Following the export ban, dozens of industry executives, entrepreneurs, and cybersecurity experts signed an open letter calling for a more scientific and transparent process for handling AI risk assessments. They argue that abruptly withdrawing powerful capabilities from defenders could be dangerous, especially as global adversaries continue to advance their own AI tools. Despite the initial national security designation, President Trump indicated during the G7 summit that negotiations with the company are progressing and that he no longer views Anthropic as a threat, though the administration’s overall regulatory strategy remains a subject of intense debate among policy experts.

The Trump administration recently implemented an export ban on Anthropic’s most advanced AI models, citing national security concerns related to a jailbreak (a method to bypass internal safety guardrails). This action forced the company to pull access to its Mythos and Fable 5 models earlier this month. The government directive, which initially gave Anthropic just 90 minutes to comply, originated after a reported disagreement with the Pentagon over modifications to guardrails for military use, leading the Department of Defense to designate Anthropic a supply chain risk. Anthropic maintains that the identified vulnerabilities are minor and comparable to those found in other models, a position supported by several independent AI safety researchers who have reviewed the government’s findings.

The incident highlights the lack of a transparent or consistent regulatory framework for AI in the United States. While the Trump administration released a national policy framework in March advocating for sector-specific regulation rather than a single rule-making body, critics describe the current enforcement approach as ad hoc and opaque. An executive order requiring companies to voluntarily share advanced models for government vetting was delayed due to concerns that it might impede innovation. Meanwhile, the administration is navigating a fragmented legal landscape, with states like California and Florida pursuing their own AI safety laws and litigation.

Anthropic, an AI firm nearing a public offering with a valuation approaching $1 trillion, remains central to these tensions. Following the export ban, dozens of industry executives, entrepreneurs, and cybersecurity experts signed an open letter calling for a more scientific and transparent process for handling AI risk assessments. They argue that abruptly withdrawing powerful capabilities from defenders could be dangerous, especially as global adversaries continue to advance their own AI tools. Despite the initial national security designation, President Trump indicated during the G7 summit that negotiations with the company are progressing and that he no longer views Anthropic as a threat, though the administration’s overall regulatory strategy remains a subject of intense debate among policy experts.