Anthropic Claude Fable 5 ban explained: Why US government restricted access for new AI models

Anthropic recently released two new AI models, Claude Fable 5 and Mythos 5, as the company expands its cybersecurity efforts. However, the US government became concerned that the models might be vulnerable to a "jailbreak" technique that could allow users to discover or exploit vulnerabilities. As a result, the government ordered Anthropic to suspend access to AI models for foreign nationals under export-control rules. 

Anthropic is now disputing the reasons behind the restrictions, saying the evidence does not justify such a broad move. However, it is complying with the government directive while seeking to have access reinstated. Now, the rollback of the powerful AI models has reignited the debate over how advanced AI systems should be regulated.

Must read: 'Don't confuse access with ownership': Sarvam CEO on what Anthropic's Fable ban means for India

What are Mythos 5 and Fable 5

Claude Mythos 5 and Claude Fable 5 are Anthropic’s most advanced, next-generation AI models. With these models, the company claims to offer "Mythos-class" capabilities that can handle long-running tasks. It is said to bring massive upgrades compared to its flagship "Opus" models.

These models are build to Mythos-class models are built to think autonomously for extended periods. Therefore, they can manage highly complex, multi-day agentic workflows without human intervention.

The Claude Fable 5 was the publicly available version of the Mythos technology, designed for general as well as enterprise use. It is built to conduct heavy-duty knowledge work, complex data analytics, and massive software engineering projects.

On the other hand, Mythos 5 was released with certain restrictions and was available to only a few trusted members. It is designed to conduct tasks like identifying deep-level automated software vulnerabilities, drug design, and biodefense screening. This model was not available for public use due to the risk of potential weaponisation.

Must read: 'Blaming Infosys, Murthy, Nandan is misplaced': Startup founder says India's real AI problem is...

Why US government ban Mythos 5 and Fable 5?

On June 12, the US government asked Anthropic to halt the rollout of Fable 5 and Mythos 5 for all foreign nationals, and it was given a short period of time to do so. Hence, access to these two models was banned outside of the US, and even for non-US citizens.

“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Anthropic models will not be affected,” Anthropic said. The company further claimed that the AI models were tested and approved by the government before releasing them to the public. Now, it is imposing restrictions because they believe the models could pose national security risks.

Anthropic highlighted that a US official discovered a narrow "jailbreak" technique that can bypass safeguards built into Fable 5. If this happens, the model could be coaxed into acting like the unrestricted Mythos 5, giving unauthorised access to powerful, automated vulnerability-discovery tools to users.

According to a Wall Street Journal report, Amazon researchers also tested the AI models and found that in some cases, the model could be manipulated to reveal information related to software security, and it was also shared with the US Department of Commerce.

Must read: 9,000,000,000 litres! That's how much water Amazon used last year

However, Anthropic suggests that the problem is not serious enough to warrant restrictions, but US officials disagree and see advanced AI as a possible national security threat.

What is Anthropic saying about the restriction?

Anthropic suggests that the government has not yet provided evidence. In the blog post, the company said, “To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws.”

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard were applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” the company added.

In addition, it highlighted that there has been a "misunderstanding" and is currently in talks with the government to commence the broader rollout of the AI models.

What does the ban mean for India?

Zoho founder Sridhar Vembu said that the restriction is a wake-up call for India. He said, “Technology is the ultimate weapon. National sovereignty, national security, all of it is now about technology.”

“Globalisation is dead, and Bharat must find her way ahead,” he added. He argued about India’s reliance on frontier models managed by global companies, and said that India must deploy smaller AI models, including Indian and Chinese open-source alternatives, to stay ahead of the race.

"With a bit of effort, we can make them work," he added. "Anyway, why pay money to people who don't even want to sell to you?"

On the other hand, Sarvam AI CEO and co-founder Pratyush Kumar said that countries should not confuse access to AI with ownership. "Fable ban is a good instigation for more people to engage in recognising the need for sovereignty," Kumar said. "But while many opine, Sarvam is at work making the Fable of India's first consequential AI company a reality."

"For AI users, it is clear that you should not confuse access with ownership, or adoption itself as advantage. And if the most significant tech differentiator you are leveraging has external control loops, then you have to accept you are vulnerable," he added.