On Friday, Anthropic announced that it is “abruptly disabling” its “most advanced AI systems for everyone”, as it has been issued a directive by the US government to suspend foreigners’ access to its products.
The order’s scope covers the company’s latest models, known as Fable 5 and Mythos 5, due to national security concerns related to possible misuse.
Regulators didn’t supply detailed written evidence, but verbally signalled the possibility of a “narrow jailbreak” approach that could circumvent certain protections implemented to stop the models from being utilised to discover software vulnerabilities, Anthropic says.
The company said it was “displeased” with the move, saying that the concerns about a small-sized exploit should not warrant the removal of commercial AI products that are already widely used.
The order is part of a more expansive policy change in the United States, which previously targeted exports of chips and is now expanding to direct access to advanced artificial intelligence systems.
There are reports that officials think frontier models may be helping to speed up more complex cyberattacks, especially against sensitive industries like banking and legacy infrastructure.
Anthropic said it had collaborated with the government on safety testing prior to introducing the models, and noted that there were comparable capabilities in other systems.
It stated that if the ruling is widely accepted, it could make the innovation of new frontier AI products in the industry much slower.
It also stated it is collaborating with authorities in order to restore access and the cloud partner AWS, adding that its other models are available.
The company has also recently filed confidentially for a US initial public offering, highlighting its growing commercial and strategic profile.
Officials, however, have defended the action as necessary for national security, underscoring an increasingly cautious approach to advanced AI deployment.
Also read: Google plans huge $40 billion investment in anthropic
