US banned Mythos and Fable for foreigners because of China concerns, report says

The US government has banned the use of Anthropic's advanced AI models, Fable 5 and Mythos 5, for foreigners globally, with the AI startup suspending access to the models for all its users. As per a new report, the White House may have made this decision to prevent China from getting access to these highly powerful models.

In Short

• US ban on Mythos 5 and Fable 5 may be related to China concerns
• Report says US feared China may jailbreak Mythos and Fable AI models
• US ban restricts access to all non-US citizens, including Anthropic employees

On Friday, the US government gave a directive to Anthropic, demanding the AI company to suspend access to its latest models, Fable 5 and Mythos 5, to all non-US citizens, inside or outside the US. While both models come with safeguards that prevent them from being used to exploit cybersecurity flaws, reports claim that the models can be jailbroken. And as per one report, US authorities feared that China may be able to exploit these flaws if it got access to it.

According to a Semafor report, the US government ordered restricted access to Fable 5 and Mythos 5 due to concerns regarding a China-linked group potentially getting access. It is unclear if China-linked groups actually did get access to the models, or whether this could have allowed the Chinese government to get their hands on both models.

If accurate, this could mark another escalation in tensions between the US and China in the AI war. While the US largely restricts sales of advanced AI chips from the likes of Nvidia to China, Beijing has previously blocked Meta’s acquisition of China-based Manus AI.

Why did White House ban Mythos 5 and Fable 5?

Mythos 5 and Fable 5 are the first Mythos-class models by Anthropic that were released to the general public. Anthropic’s Mythos family of models is said to be so advanced that it can potentially hack any software out there by identifying flaws humans didn’t know existed.

Anthropic did implement sufficient safeguards on Mythos and Fable that would stop the AI models from giving responses to any sensitive topics such as cybersecurity. However, as per reports, Amazon researchers were able to jailbreak the models. They reportedly used prompts that led the model to reveal information related to software vulnerabilities that was meant to remain restricted.

That is, if correct prompts were used, anyone with access to Mythos 5 or Fable 5 could, in theory, be able to potentially plan cyberattacks with these models. Do note that previously, it was reported that the US National Security Agency (NSA) may be preparing to use Mythos to plan cyberattacks, with help from Anthropic engineers.

Why is White House concerned about China?

It is likely that US authorities may have feared that China’s access to these models could create a national security risk. The report states that China could have also tried to copy the models via distillation – training smaller models on Fable 5 or Mythos 5.

Though, as per the report, an Anthropic spokesperson stated that the White House didn’t raise Chinese access to Mythos in its conversations around the Fable jailbreak and export controls.

Keep in mind that the US’ ban on access for foreigners was not limited to consumers. As part of the ban, even Anthropic employees who are not US citizens are not allowed to use the two models.

However, as per a report from Axios, Anthropic has sent its engineers to Washington DC to hold talks with US officials to find a solution and restore access to both AI models.

- Ends