Executives from more than 50 cybersecurity organizations have urged the U.S. government to reconsider recent restrictions on Anthropic’s most capable AI models, arguing that the move could undermine efforts to detect and fix software vulnerabilities.
The appeal comes after Washington ordered that Anthropic suspend foreign national access to its Fable 5 and Mythos 5 models on national security grounds. Industry leaders told policymakers the decision will limit cybersecurity teams as other AI tools increasingly enable malicious actors to identify and exploit software weaknesses.
Anthropic has previously acknowledged the hacking potential of its Mythos model and withheld it from broad release to reduce the risk of misuse. The company later published a public variant called Fable, which it described as incorporating safeguards designed to prevent use in identifying software vulnerabilities.
Signatories to the letter said Anthropic’s models are not uniquely capable of discovering and weaponizing vulnerabilities. The letter cited competing systems, including an open-source Chinese model identified as Kimi 2.7, as offering similar functionality. In that context, several cybersecurity leaders argued that removing access to Anthropic’s capabilities from a wide group of researchers and practitioners would be counterproductive.
Joshua Saxe, chief technology officer at AI security firm Abundant Security and a signer of the letter, characterized Mythos as likely the leading model for finding security bugs and code issues, but framed it as an incremental improvement over other already open models. "Mythos is almost definitely the best model right now for finding security bugs and codes, but it is like an incremental advance over other models that are already open," he said in an interview.
Senior Anthropic staff are scheduled to meet with government officials at the U.S. Department of Commerce in Washington on Monday, according to an administration official. Anthropic has said the government believes there is a method to bypass, or "jailbreak," a safeguard that prevents Fable from being used to identify software vulnerabilities. The company has argued that a narrow potential jailbreak should not justify cutting off access to a model used by hundreds of millions of people.
The industry letter reinforced Anthropic’s claim that the company already implemented robust protections, and warned that removing the models from the hands of defensive researchers could be "dangerous". Signatories noted that China’s open-source models are only months behind the most capable U.S. systems and suggested Beijing may possess capabilities not visible in public releases.
Alex Stamos, chief product officer at Corridor and a signatory, criticized the government action as lacking clarity and consistent application. He said any regulation should be evidence-based and clearly defined, and asserted those standards were not met in this instance. Stamos described the move as "an overreaction by the government," and referenced a dispute between Anthropic and the third party that flagged the issue regarding the seriousness of the findings, based on his conversations with those involved.
Separately, cybersecurity firm CrowdStrike said last week that China-linked hacking groups represented the top espionage threat to technology companies over the past year. The letter from industry leaders and the administration’s restrictions intersect with growing concern about state-aligned and criminal actors using AI-enabled tooling to accelerate offensive campaigns.
In recent months the $965 billion AI company, which is preparing to go public, has previously tussled with the U.S. government over access to its models and their implications for national security. Earlier in the year, the Trump administration directed U.S. agencies to stop working with Anthropic and labeled it a supply risk, citing the company’s reluctance to allow its technology to be used for mass surveillance and autonomous weapons.
As the government and industry prepare for a meeting at the Commerce Department, the debate centers on how to balance the risks of misuse against the practical needs of cybersecurity practitioners who rely on advanced tools to find and repair software flaws. The signatories contend that restricting access to defensive uses of such models may hamper the broader effort to stay ahead of attackers who can repurpose similar AI capabilities.
Context and next steps
The scheduled discussions at the Commerce Department will be a focal point for clarifying the government’s specific concerns and the evidence behind the decision to limit access. Industry leaders who signed the letter are pressing for regulation that is narrowly tailored, clearly defined and consistently applied, while warning against an outcome that would reduce defensive visibility into emergent threats.