A growing group of cybersecurity professionals is urging the United States government to reconsider its recent restrictions on Anthropic’s advanced AI models, arguing that the decision could weaken digital defenses at a time when cyber threats are rapidly evolving.

More than 70 security experts have signed an open letter calling for the removal of export controls placed on Anthropic’s powerful AI models, Fable and Mythos. According to the signatories, limiting access to these models could make it harder for cybersecurity teams to identify vulnerabilities, secure software, and defend critical systems against increasingly sophisticated attacks.

Why the US Government Restricted the Models

Last week, the U.S. government ordered Anthropic to limit the export of its Fable and Mythos AI models, citing national security concerns. While officials did not publicly explain the exact reasons behind the decision, Anthropic responded by suspending access to the models worldwide.

The move immediately sparked debate across the cybersecurity industry, particularly because Mythos was specifically designed to assist with vulnerability discovery and software security research.

Anthropic originally introduced Mythos as a restricted-access cybersecurity model capable of finding software vulnerabilities at an advanced level. Access was initially limited to a small number of organizations before gradually expanding to selected security firms and research groups.

Fable, released later as a public version of Mythos, included strict safeguards intended to prevent misuse in cybersecurity, biology, and chemistry-related activities.

Security Experts Say Defenders Need Advanced AI Tools

The cybersecurity community argues that restricting these AI models hurts defenders more than attackers.

In their open letter, security researchers warned that removing advanced AI capabilities from legitimate cybersecurity professionals could leave organizations less prepared to identify and fix vulnerabilities before malicious actors exploit them.

The letter states that cybersecurity teams increasingly rely on AI tools to automate tasks such as:

  • Identifying software vulnerabilities
  • Reviewing source code
  • Testing security patches
  • Detecting configuration weaknesses
  • Improving overall software resilience

Experts argue that these activities are essential for protecting businesses, governments, and critical infrastructure.

Industry Veterans Join the Protest

The open letter has attracted support from some of the most respected names in cybersecurity.

Among the signatories are former security executives, cryptographers, bug bounty leaders, vulnerability researchers, and security company founders who believe the restrictions could unintentionally weaken defensive security efforts.

Many of these experts have spent decades helping organizations defend against cyberattacks and view AI as a powerful tool for improving software security rather than creating additional risks.

The Debate Over AI Guardrails

Part of the controversy stems from concerns surrounding a research paper that reportedly demonstrated ways to bypass some of Fable’s built-in safety restrictions.

According to cybersecurity experts familiar with the research, the technique did not represent a true security bypass. Instead, researchers reportedly asked the AI model to identify and fix known vulnerabilities within software code — a task that many security professionals consider routine defensive work.

Critics argue that if AI models cannot assist with finding and fixing software flaws, they lose much of their value for cybersecurity teams.

Supporters of the restrictions, however, believe that highly capable cybersecurity-focused AI systems could potentially be misused by threat actors if adequate safeguards are not in place.

Similar Capabilities Exist Elsewhere

Another key argument raised by security researchers is that similar vulnerability analysis capabilities already exist in other advanced AI models.

According to the open letter, many of the functions demonstrated using Fable can also be performed by leading AI systems from other providers, including models from OpenAI and several international AI companies.

Researchers claim that restricting one company’s models may do little to prevent misuse while limiting access for legitimate defenders who rely on AI-powered security assistance.

Balancing Innovation and National Security

The situation highlights a growing challenge facing governments around the world: how to regulate powerful artificial intelligence systems without slowing innovation or weakening cybersecurity defenses.

As AI becomes increasingly capable, policymakers must balance national security concerns with the practical needs of researchers, software developers, and security teams.

Many experts believe that future regulations should be developed through transparent collaboration between governments, academia, and industry leaders rather than through emergency restrictions.

What Happens Next?

At the moment, Anthropic’s Fable and Mythos models remain unavailable due to the export control order.

The cybersecurity community continues to push for a reassessment of the decision, arguing that advanced AI tools can play a critical role in defending against cyber threats when used responsibly.

As governments and technology companies work to define the future of AI regulation, this debate may become one of the most important examples of how society balances innovation, security, and access to powerful emerging technologies.

For now, the discussion surrounding Anthropic’s AI models serves as a reminder that the future of cybersecurity may depend as much on policy decisions as it does on technological breakthroughs.

Read More on VitalStack

Enjoyed this article?

Subscribe for weekly deep-dives on AI and health — straight to your inbox.