The United States government has abruptly ordered artificial intelligence company Anthropic to block foreign nationals both inside and outside the country from accessing its most advanced AI systems, marking one of the strongest government interventions yet in the rapidly escalating global race to control frontier artificial intelligence.
The emergency directive, issued by the U.S. Commerce Department, forces Anthropic to immediately disable access to its flagship AI models, Fable 5 and Mythos 5, after federal authorities raised alarm over serious national security vulnerabilities linked to the technology.
The sudden move effectively triggered a total shutdown of these models worldwide, with Anthropic now scrambling to comply while warning that similar policies could paralyze innovation across the entire AI industry.
According to reports, the Commerce Department enacted export control restrictions directly targeting Anthropic’s most advanced AI systems after identifying what officials described as a dangerous security vulnerability capable of allowing users to bypass built-in safeguards.
The government specifically ordered the company to block access for all foreign nationals, including non-U.S. citizens currently living inside the United States.
The restrictions also extend globally, affecting international users accessing Anthropic systems outside American borders.
U.S. Commerce Secretary Howard Lutnick confirmed the emergency directive while explaining the scope of the restrictions.
“These models are now subject to export restrictions, which prohibits usage by customers outside the U.S. and foreign nationals within the U.S.”
The order immediately forced Anthropic to suspend normal operations connected to both affected models.
The restrictions apply specifically to Anthropic’s newest frontier AI systems (Fable 5 and Mythos 5) considered among the company’s most advanced artificial intelligence products currently available.
Federal officials reportedly grew concerned after internal security investigations identified a possible “jailbreak” technique capable of overriding the models’ safety barriers.
In artificial intelligence systems, jailbreaking refers to methods that manipulate an AI model into ignoring restrictions designed to prevent harmful or dangerous outputs.
Government investigators reportedly determined malicious actors could potentially exploit the discovered vulnerability to force the AI systems into performing dangerous tasks beyond intended safety boundaries.
As a result, Washington moved quickly to contain the risk before broader exploitation became possible.
The strongest fears inside the Commerce Department center around advanced cybersecurity threats.
Officials reportedly believe bad actors who successfully bypass the AI models’ built-in safeguards could weaponize the systems to identify software vulnerabilities at scale.
Security experts fear the AI could potentially conduct highly sophisticated agentic hacking, where artificial intelligence autonomously scans digital infrastructure, identifies weaknesses, and assists in launching cyberattacks with minimal human involvement.
Government analysts worry such capabilities could target critical infrastructure systems including banking networks, healthcare databases, telecommunications systems, and national power grids.
Testing also showed the underlying Mythos 5 model possessed highly advanced biological reasoning capabilities capable of solving complex scientific challenges involving advanced virus development.
Federal officials fear that without strict safeguards, highly capable frontier AI systems could assist malicious actors in researching or designing dangerous biological agents.
The possibility that advanced AI models could accelerate bioweapons development has increasingly become one of the most serious concerns surrounding next-generation artificial intelligence systems globally.
Another major concern involves what experts call AI distillation.
Distillation occurs when developers systematically interact with a powerful AI system millions of times in order to reverse engineer how the model thinks and replicate its reasoning capabilities on separate infrastructure.
U.S. intelligence agencies worry foreign adversaries could exploit Anthropic’s advanced models to effectively steal American AI breakthroughs.
If successful, rival governments could recreate similarly powerful AI systems on their own servers without American oversight, safety controls, or regulatory restrictions.
Washington increasingly views advanced AI models as strategic technologies requiring protection similar to classified defense systems.
Anthropic has openly challenged how the U.S. government handled the emergency directive.
The company released a sharply worded statement criticizing what it described as an opaque process lacking technical transparency.
“As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles,” the company said.
Anthropic also warned that applying similar standards broadly across the AI industry could severely damage innovation.
The company argued that minor or repeatable security bypasses should not automatically justify complete shutdowns.
Because the government order requires strict citizenship verification standards, Anthropic says compliance has become technically impossible without disabling access far more broadly.
The company has reportedly instructed cloud infrastructure partners including Amazon Web Services to revoke model access across all regions while compliance systems are implemented.
Anthropic warned regulators that extending similar restrictions industry-wide could halt deployment of new frontier AI models entirely.
