- Anthropic's AI models Fable 5 and Mythos 5 access has been globally restored after US approval
- Mythos 5 excels at finding and fixing computer system weaknesses but was initially restricted
- US banned non-citizen access after a minor jailbreak issue was found in Fable 5 soon after launch
After a US government greenlight, global access to Anthropic's powerful AI model Fable 5 -- along with restricted access to the even more powerful Mythos 5 has been restored, following a standoff that strained ties between Washington and Silicon Valley.
What Is Mythos?
Mythos is the name of the most advanced series of AI models made by Anthropic, the San Francisco company behind the Claude chatbot. Mythos is especially good at finding weaknesses in computer systems and at fixing them.
Because that skill could be used by hackers, Anthropic did not release the full version to the public at first. Only a small group of trusted partners first in the United States only, then internationally could use it.
On June 9, Anthropic released a public, safer version called Fable 5. It had safety filters built in to stop people from misusing its hacking skills.
How Did The Export Ban Happen?
Just three days after Fable 5 launched, the government ordered Anthropic to cut off access to both Fable 5 and Mythos 5 for anyone who was not a US citizen, even Anthropic's own foreign employees.
Since Anthropic could not easily check users' nationality, it shut down Fable 5 and Mythos 5 for everyone, everywhere.
The government said someone had found a way to "jailbreak," or trick, Fable 5 into giving dangerous information. Anthropic said the problem was minor and that other AI models could do the same thing.
Why The U-Turn?
Before Mythos's arrival, the Trump administration wanted fewer rules on AI companies, not more, hoping that would help the US beat China in the AI race.
But Mythos came onto the scene when Trump's relationship with Anthropic was already rocky.
In February, Anthropic, which has ties to Democrats and pushes a pro-safety stance on AI, refused to let the Pentagon use its AI without limits on surveillance and weapons.
Trump in response ordered every government agency to stop using Anthropic's products, calling the company "Leftwing nut jobs" on social media.
Some observers believe the government's decision to curb the release of Mythos 5 and Fable 5 was politically motivated and connected to this earlier spat, not AI safety.
Anthropic archrival OpenAI, the ChatGPT-maker that enjoys smoother ties with the Trump White House, was asked to restrict the release of its own Mythos-equivalent technology, GPT-5.6 to a limited set of approved partners.
Is The White House Divided?
Speaking on his company's podcast, White House adviser Marc Andreessen, a major venture capitalist, said the administration is actually torn between two goals that are hard to have at the same time.
One group wants to spread American AI as widely as possible around the world, so that the US -- not China -- sets the rules for how AI is used everywhere.
The other group wants to do the opposite: hold back the most powerful AI models out of fear they could be used to hack banks, hospitals or government systems, including by China.
Andreessen, who supports deploying AI instead of curbing it, admitted his own side of the argument is losing, suggesting the administration is leaning toward more restriction rather than the wide release he favors.
Where Do Things Stand Now?
After easing the ban on Mythos last week to re-allow access to a small group of American cybersecurity firms, the US government told Anthropic on Tuesday that it could also resume global access to Fable 5.
Still, for now, the rules of the road in the US for AI companies remain opaque, with the fate of international users of Mythos 5, which include some major institutions like the European Commission, still undefined.
This could change in August. In accordance with an executive order from the White House, the government is drawing up criteria for which models would fall under new security restrictions.
Anthropic meanwhile called for standards to both assess critical vulnerabilities in advanced models and respond to them, and said it will work with Amazon, Microsoft, Google and others on the effort.
(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated feed.)
