More than 50 cybersecurity professionals have publicly called on the US government to lift export controls restricting access to Anthropic’s latest frontier large language models, warning that the move risks weakening defenders while adversaries continue to advance.

The open letter, addressed to US Secretary of Commerce Howard Lutnick and US National Cyber Director Sean Cairncross, follows a June 12 announcement from Anthropic that the US government had issued an export control directive suspending access to Fable 5 and Mythos 5 by any foreign national.

Anthropic said the directive forced it to suspend access to both models for all customers to ensure compliance. The models had been released only days earlier.

Fable 5 was presented by Anthropic as a general-access LLM powered by the same underlying frontier AI model as Mythos 5, which was itself an upgrade from Claude Mythos Preview, but with additional guardrails. The company said those guardrails were particularly important in areas such as cybersecurity, where the technology “could be misused to cause serious damage.”

The US government invoked “national security concerns” to explain the directive. Anthropic said it believes the action originated from research that allegedly found a method of bypassing Fable 5’s guardrails.

“We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple and we have found that other publicly-available models are able to discover them as well without requiring a bypass,” stated Anthropic.

The company also denied the existence of a “universal jailbreak” for Fable 5.

Cyber community speaks out 

In the open letter, the cybersecurity professionals asked the US government to reverse the restrictions and adopt a more transparent process for assessing AI risks.

“We, the undersigned executives and technical leaders from across the United States and its allies, write to you to ask you to lift the export control directives on Anthropic’s Fable and Mythos large language models and commit to an open, scientific and transparent process of handling AI risk assessments in the future,” the letter said.

The signatories argued that AI is already having a major effect on cybersecurity, including by lowering the difficulty of finding software flaws and writing exploits. They also acknowledged that Anthropic’s Mythos-class models are effective at identifying vulnerabilities and weaponising exploits.

However, the letter said the models are not unique in this respect, and that many cybersecurity professionals already use other foundation and open-source models for security audits and red-teaming.

“Anthropic’s Mythos-class models are quite good at finding flaws and weaponizing exploits,” the letter said. “However, they are not uniquely good at these tasks, and many of the undersigned individuals regularly use other foundation and open-source models for security audits and red-teaming every day.”

The letter also said Anthropic had built protections into Fable to prevent cyber-offensive use, arguing that these restrictions were already notable within the security community.

“Anthropic has built multiple protections into the Fable model to prevent its use for cyber offensive uses. These protections were so aggressive as to be the source of humor in the cyber community on launch day,” the signatories said.

A central concern raised in the letter is that removing access to advanced AI tools could leave legitimate security teams at a disadvantage. The signatories argued that access to AI is increasingly important for developers and defenders seeking to identify and fix vulnerabilities in both new and legacy code.

“It is essential to provide AI to coders and security teams so they can find and fix flaws in their own newly-written as well as decades of legacy code faster than our adversaries,” the letter said.

The signatories also warned that Chinese open-weight models are “only months behind the best American models,” adding that “it seems likely that the PRC government has access to private capabilities beyond what has been published.”

“To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous,” they said.

The letter further challenged the apparent basis for the directive, saying the underlying research focused on whether a human-prompted section of code was insecure. The signatories argued this is a necessary capability for any model intended to write secure code, and should not automatically be considered offensive.

“Were focused on determining whether a human-prompted section of code was insecure. This is a necessary capability in any model that is intended to write secure code and should not be considered an offensive capability,” the letter said.

The signatories also disputed the idea that Fable provides a unique uplift in cybersecurity capability compared with other AI systems.

“Can be replicated on GPT-5.5, Opus, Sonnet and even Chinese models like Kimi 2.7. The justification for this unprecedented action was that Fable provides a unique “uplift” of capabilities beyond other AI models, but AI has been finding bugs and generating working exploits at superhuman levels since last year,” the letter said.

They added that Anthropic is already addressing the research, and that security findings should be used to improve systems rather than justify broad bans.

“Anthropic is addressing the research. As security professionals, we recognize that our work does not lead to a simple end-state where a system is fully safe, and the purpose of research like this is to enable continuous improvement, not to ban the technology,” the letter said.

Recommended reading

• Anthropic Investigates Claimed Unauthorised Access to Mythos
• Anthropic’s Mythos Lands It a White House Meeting
• Is Mythos as Dangerous as Anthropic Claims?

The letter claimed the directive has removed important tools from defenders, created uncertainty in the market, and potentially damaged US leadership in AI.

“As a result, this action has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it,” the signatories said.

While the cybersecurity professionals said they do not all agree on whether AI regulation is the right approach, they argued that any model restrictions should be grounded in scientific evaluation, developed with input from industry and academia, created through democratic rule-making, enforced transparently and fairly, and used only to the minimum extent needed to protect the public.

“Not all of us agree that AI regulation is the right way forward. But if this Administration’s laudable goal of securing our nation’s critical infrastructure is going to include models being regulated, then the regulations should be:

1. Grounded in scientific evaluations developed with input from industry and academia;
2. Created through a democratic rule-making process;
3. Enforced transparently and fairly with appropriate time given to remediate; and
4. Used only to the minimal extent necessary to ensure the safety of the American public,” the letter said.

Signatories named on the letter include Alex Stamos, Chief Product Officer at Corridor; Feross Aboukhadijeh, CEO of Socket; Abhishek Arya, Principal Engineer at Google and founder of OSS-Fuzz; Kevin Bankston, Senior AI Governance Advisor at the Center for Democracy & Technology; Anthony Bettini, CEO of VulnCheck; Casey John Ellis, founder of disclose.io and Bugcrowd; Matthew D. Green, Associate Professor at Johns Hopkins University; Mikko Hypponen, CRO at Sensofusion Finland; Joe Levy, CEO of Sophos; Katie Moussouris, CEO of Luta Security; Riana Pfefferkorn, Policy Fellow at Stanford HAI; Bruce Schneier of Harvard University and the University of Toronto; Eugene H. Spafford, Distinguished Professor at Purdue University; Rachel Tobac, CEO of SocialProof Security; and Chris Wysopal, co-founder of Veracode.

The letter states that affiliations are included for reference only and do not indicate organisational endorsement.