Cyber leaders urge US to lift curbs on Anthropic’s security models

More than 50 cybersecurity industry leaders have signed an open letter to the Trump administration demanding it reverse emergency export controls slapped on two of Anthropic’s most advanced AI models. Their argument is straightforward: the restrictions are handicapping the people trying to stop cyberattacks while doing essentially nothing to slow down the people launching them.

The letter, dated June 15, 2026, landed just two days after the US Commerce Department issued an emergency order requiring Anthropic to block foreign nationals from accessing its Fable 5 and Mythos 5 models. The trigger was a reported jailbreak of cybersecurity safeguards built into the models, which the government deemed a national security concern serious enough to warrant immediate action.

What happened and why it matters

Here’s the timeline. Anthropic released its Mythos-class models, designed for both offensive and defensive cyber applications, to a limited audience around June 9-10. Within days, someone reportedly found a way around the models’ safety guardrails. By June 13, Commerce had issued its emergency order. By June 15, the cybersecurity industry was in open revolt against it.

The signatories include executives from Nvidia, Adobe, and other major technology firms, alongside prominent cybersecurity figures like Joshua Saxe and Alex Stamos. Their core claim is that the restricted models don’t actually provide capabilities beyond what’s already available through other platforms, including open-source alternatives.

Anthropic’s response to the Commerce Department order was to globally suspend access to both advanced models. The company determined that selectively enforcing restrictions based on nationality was impractical, so it chose the nuclear option of shutting everyone out. That means American cybersecurity teams lost access too, which is precisely the outcome the letter’s signatories are furious about.

The bigger picture: Anthropic vs. Washington

This isn’t an isolated dustup. The export controls landed amid growing friction between Anthropic and the US government across multiple fronts, including model safety standards, military applications, and a prior designation as a supply-chain risk by the Department of Defense.

The DoD supply-chain risk designation suggests a deeper institutional skepticism toward Anthropic that predates the jailbreak incident. The company has previously taken public stances against enabling mass surveillance or military uses of its technology.

Anthropic has been valued near $965 billion and is reportedly preparing for an IPO. The timing of these restrictions could hardly be worse from a business perspective.

The broader context here is Project Glasswing, an initiative aimed at enhancing cybersecurity capabilities using advanced AI. The export controls create an awkward contradiction: the government wants better AI-powered cyber defense but is simultaneously restricting access to some of the most capable models built for exactly that purpose.

What this means for investors

For Anthropic itself, the restrictions introduce regulatory risk at a critical moment in the company’s trajectory toward a public listing. A near-trillion-dollar valuation assumes continued access to markets and customers.

Open-source AI models could be indirect beneficiaries. If the letter’s signatories are correct that comparable capabilities exist in open-source alternatives, then the export controls may accelerate adoption of those alternatives.

For Nvidia and Adobe, whose executives signed the letter, the concern is partly about their own ecosystems. Both companies have deep integrations with AI model providers, and restrictions that limit the deployment of advanced models could slow adoption of their hardware and software platforms.

The lobbying dynamic here is also worth watching. Over 50 signatories from major firms represents a significant concentration of industry pressure. If the administration holds firm, it signals a willingness to prioritize national security concerns over industry objections, which could set precedents for how other advanced AI models are regulated. If it caves, it establishes that coordinated industry pushback can reverse emergency national security orders in a matter of days.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.