US Cyber Experts Urge Lifting of Ban on Anthropic’s AI Models

Over 50 cybersecurity professionals have signed a public appeal urging the US government to reverse its ban on Anthropic’s newest large language models, Mythos 5 and Fable 5. The group argues that restricting access to these frontier AI systems does more harm than good, especially when adversaries are already advancing their own capabilities.

On June 12, Anthropic disclosed that the US government had issued an export control directive blocking all foreign nationals from accessing Fable 5 and Mythos 5, which had only been released days earlier. To comply, the company suspended access to both models for all customers. Fable 5 was positioned as a general-access LLM built on the same underlying technology as Mythos 5, an upgrade from Claude Mythos Preview, but with additional guardrails. Anthropic specifically noted that these guardrails targeted areas like cybersecurity, where the model “could be misused to cause serious damage.”

The US government cited national security concerns as the basis for its directive. Anthropic believes the decision stemmed from research that allegedly demonstrated a method to bypass Fable 5’s guardrails. However, the company pushed back on that claim. “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,” Anthropic stated. The company denied the existence of a “universal jailbreak” for Fable 5.

Two days later, a coalition of 54 CISOs, cybersecurity practitioners, and vendors sent an open letter to US Secretary of Commerce Howard Lutnick and National Cyber Director Sean Cairncross. The group requested that the export controls on Fable 5 and Mythos 5 be lifted. They also called on the government to “commit to an open, scientific and transparent process of handling AI risk assessments in the future.”

While the signatories acknowledged that Anthropic’s latest models are “quite good at finding flaws and weaponizing exploits,” they stressed that these are not the only tools capable of such tasks. The ability to identify insecure code is a fundamental feature of any secure coding assistant, they noted, pointing to equivalent capabilities in other models such as OpenAI’s GPT-5.5, Anthropic’s own Claude Opus and Sonnet, and Chinese models like Kimi 2.7.

The group also recognized Anthropic’s efforts to prevent Fable from being used for “cyber offensive uses” and noted that the company is now addressing the research that likely triggered the US government’s decision. “To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous,” the letter warned. It further argued that the government’s action has created market uncertainty and risks undermining America’s AI leadership “without any real risk to justify it.”

Notable signatories included Alex Stamos, chief product officer at Corridor and former chief security officer at Facebook and Yahoo; Joe Levy, CEO of Sophos; and Sandra McLeod, CISO at Zoom Communications. Their position has resonated with other cybersecurity experts. William Wright, CEO of Closed Door Security, did not sign the letter but shared a similar view. He said the US reaction “suggests that the worries around jailbreaking these models are real,” but banning access is the wrong approach. “Cutting off access to the model so abruptly will cause huge logistical problems, both within Anthropic and within any critical industry partners given access to the model. Rather than foster resilience, this move creates chaos,” he explained. Wright urged the government to work “transparently and with clear guiding principles” alongside AI and cybersecurity experts.