Over 100 Cyber Leaders Urge Federal Agencies to Allow Fable Access

Post Views: 9

Over 100 cybersecurity professionals and industry leaders have released an open letter urging the U.S. government to rescind an export control order that shut down Anthropic’s Fable 5 and Mythos 5 models.

Government Action and Anthropic’s Response

The directive, issued by federal authorities, required Anthropic to restrict access to these AI systems for all users, regardless of location. Anthropic complied with the measure, citing the need to adhere to regulatory requirements. The government cited national security concerns, referencing a specific vulnerability in Fable 5 that allowed users to analyze codebases and identify software flaws.

Anthropic’s Argument

Anthropic described this as a limited, non-universal exploit and argued that similar functionalities exist in other models such as OpenAI’s GPT-5.5. The company emphasized that restricting access to advanced AI tools undermines cybersecurity efforts and creates an imbalance in defensive capabilities.

Cybersecurity Coalition’s Open Letter

A coalition of cybersecurity experts, led by Alex Stamos, former chief security officer at a major tech firm, published an open letter titled “On Transparent AI Cyber Protections” addressed to U.S. Secretary of Commerce Howard Lutnick and National Cyber Director Sean Cairncross. The document, hosted on freefable.org, contends that Mythos-class models, while effective, are not uniquely capable of identifying and exploiting security weaknesses.

Key Claims in the Letter

The letter argues that the specific exploit triggering the government’s action only enabled secure code review, a function already available in other AI systems. It highlights that similar capabilities exist in models like Anthropic’s Opus and Sonnet, as well as the open-source Kimi 2.7 developed by Moonshot. The document stresses the importance of providing AI tools to developers and security teams to accelerate vulnerability detection and remediation.

Signatories and Industry Endorsements

The letter has garnered 126 endorsements from industry leaders, including 26 former and current chief information security officers (CISOs), 21 CEOs, and representatives from research institutions, startups, and technology firms. Notable signatories include Sophos CEO Joe Levy, CISO Ross McKerchar, Socket CEO Feross Aboukhadijeh, and former NSA official Vinh Nguyen.

Industry Perspectives

Levy emphasized that defenders require the most advanced tools to counteract adversarial threats, stating that the most capable AI models are critical for rapid vulnerability resolution. Critics of the letter, including security expert Davi Ottenheimer, have challenged its portrayal of Mythos’ offensive capabilities, arguing that previous assessments framed these models as significant threats.

Debate Over AI Regulations

Zack Korman, cofounder of AI security firm Embroidery, criticized the letter for contradictory claims, noting that even if restrictions were lifted, most defenders would still lack access to these tools due to Project Glasswing’s limited availability. The debate underscores broader tensions between national security priorities and the need for open AI innovation.

Government and Industry Tensions

The government maintains that export controls are necessary to prevent adversarial use of advanced AI, while industry leaders argue that such measures hinder collaborative efforts to strengthen global cybersecurity defenses. The situation remains under close scrutiny as stakeholders continue to navigate the complex interplay between regulatory oversight and technological progress.

The letter advocates for AI regulations grounded in scientific evidence, democratic processes, and transparent enforcement, with restrictions applied only when absolutely necessary.