More than 120 cybersecurity and technology professionals have formally petitioned the U.S. federal government to reverse its recent export control directives on Anthropic’s Fable and Mythos large language models. The open letter, dated Sunday (June 14), urges Commerce Secretary Howard W. Lutnick and National Cyber Director Sean Cairncross to lift restrictions that have suspended access to these advanced AI tools, arguing the move jeopardizes American AI leadership and hinders cybersecurity defenders.
Government Directive Triggers Access Suspension
The appeal follows Anthropic’s announcement on Friday (June 13) that it had disabled access to its Fable 5 and Mythos 5 AI models. This action was a direct response to a U.S. government export control directive, which mandated the company to suspend access for ‘any foreign national,’ a broad category that includes Anthropic employees, whether they are located within or outside the United States. The models, launched just four days prior on Tuesday (June 9), were initially presented by Anthropic with built-in safeguards designed to prevent their misuse in sensitive areas such as cybersecurity, biology, and chemistry, as well as distillation processes.
Tech Professionals Unite Against Restrictions
The open letter, which has garnered signatures from over 120 cybersecurity and other tech professionals, was publicly shared by Joe Levy, CEO of cybersecurity firm Sophos, in a LinkedIn post on Monday (June 15). Levy noted that he and numerous colleagues from across the security community had endorsed the letter. The signatories collectively advocate for a new, industry-inclusive process for assessing artificial intelligence risk, emphasizing that the current approach is counterproductive to national security interests and technological advancement.
Experts’ Core Arguments: AI’s Dual Role and Global Race
Central to the experts’ argument is the belief that AI is already exerting significant impacts on the cybersecurity landscape, fundamentally altering both offensive and defensive capabilities. While acknowledging the prowess of Anthropic’s Mythos models in identifying software flaws, the letter asserts that these models are ‘not uniquely good at it,’ implying that similar capabilities are either already present in other models or can be rapidly developed. Furthermore, the letter highlights Anthropic’s proactive measures, stating explicitly that the company ‘has built protections into the Fable model to prevent its use for cyber offensive purposes,’ suggesting a responsible approach to development. The professionals underscore the critical need to provide advanced AI tools to coders and security teams, emphasizing that withholding such tools from defenders creates a strategic disadvantage. They issue a stark warning regarding global AI competitiveness, noting that ‘Chinese open-weight models are only months behind the best American models.’ This rapid global advancement, they argue, creates a dangerous scenario where ‘it is dangerous to keep the best capabilities from defenders when adversaries are making rapid advances,’ thereby undermining national security rather than enhancing it. The letter posits that access to cutting-edge AI is paramount for maintaining a robust defensive posture in an increasingly complex digital threat environment.
Market Uncertainty and Risk to U.S. AI Leadership
The open letter directly challenges the underlying premise and efficacy of the government’s action on Fable and Mythos. Signatories contend that the specific capability within the models that triggered the export control ‘should not be considered an offensive capability,’ suggesting a mischaracterization of the technology’s primary utility. They further argue that Fable’s capability to find bugs and generate working exploits is not exclusive to Anthropic’s models, stating it ‘can be replicated by other models,’ thus diminishing the perceived unique risk. The experts also point to Anthropic’s ongoing commitment to iterative improvement, noting that the company ‘is addressing research to enable continuous improvement,’ indicating a dynamic and evolving development process that can address concerns. The immediate and broader consequences of the directive, according to the open letter, are severe: it ‘has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it.’ This assessment paints a picture of unintended negative consequences, not only for U.S. technological competitiveness but also for the practical capabilities of cybersecurity professionals tasked with protecting critical infrastructure and data.
Proposed Framework for AI Regulation
While acknowledging that ‘not all signers agree that AI regulation is the way to go,’ the letter offers a comprehensive framework for any future regulatory measures. Should regulation be implemented, the experts propose that it must incorporate input from both industry and academia, ensuring a comprehensive and informed perspective. They advocate for a ‘democratic rule-making process’ to foster transparency and broad consensus. Furthermore, any enforcement of such regulations should be conducted ‘transparently and fairly,’ and applied ‘only to the extent necessary to protect the American public.’ The letter concludes this section by suggesting that adherence to these principles would enable the federal government and industry to ‘partner to help maintain America’s lead in technology while protecting critical software and systems,’ fostering innovation while safeguarding national interests.
The open letter represents a significant pushback from the tech community against what they perceive as an overly restrictive and potentially damaging government intervention in AI development. As Anthropic reportedly works with the White House to resolve the restrictions on its Fable and Mythos models, the debate highlights the complex balance between national security concerns and the imperative to foster innovation and maintain technological leadership in the rapidly evolving field of artificial intelligence.
