The White House is preparing to give federal agencies access to Anthropic's Mythos AI model, a powerful tool that has identified thousands of cybersecurity vulnerabilities but also raises concerns about potential misuse.
Gregory Barbaccia, federal chief information officer at the White House Office of Management and Budget, notified Cabinet department officials this week that his office was establishing protections to enable agency use of the AI system. The move comes as part of Anthropic's controlled "Project Glasswing" initiative, which allows select organizations to test the unreleased Claude Mythos Preview model for defensive cybersecurity purposes.
We're working closely with model providers, other industry partners, and the intelligence community to ensure the appropriate guardrails and safeguards are in place before potentially releasing a modified version of the model to agencies
Gregory Barbaccia, Federal CIO — Bloomberg News
Mythos represents a significant leap in AI-powered cybersecurity capabilities. The model has demonstrated an unprecedented ability to identify vulnerabilities in operating systems, web browsers, and other critical software infrastructure. Its advanced coding capabilities enable it not only to detect security flaws but also to understand how they might be exploited.
The deployment timeline remains unclear. Barbaccia's email, with the subject line "Mythos Model Access," did not specify when agencies might gain access or provide details about implementation protocols. The cautious approach reflects the dual-edged nature of the technology — while it offers powerful defensive capabilities, the same features that make it effective at finding vulnerabilities could theoretically be used for offensive purposes.
The Indian outlet frames this as a straightforward technology deployment story, emphasizing the technical capabilities and government process without dwelling on geopolitical implications. Their coverage reflects India's position as a major tech hub interested in AI developments but not directly involved in US cybersecurity policy decisions.
The Singaporean outlet presents a factual account focused on the administrative and technical aspects of the deployment, reflecting Singapore's role as a regional tech and financial hub that closely monitors US technology policy developments. Their framing avoids taking sides while highlighting the security implications relevant to their own cybersecurity-conscious audience.
Anthropic announced Mythos on April 7 as part of a carefully managed rollout designed to balance innovation with security concerns. The company has been in discussions with the Trump administration about the technology, according to co-founder Jack Clark, even as the Pentagon recently severed business ties with the AI lab following a contract dispute.
The federal government's interest in Mythos underscores the growing recognition that AI systems will play a crucial role in national cybersecurity infrastructure. As cyber threats become more sophisticated, traditional security tools struggle to keep pace with the volume and complexity of potential vulnerabilities in modern software systems.
The initiative raises questions about how the government will balance the benefits of advanced AI capabilities against the risks of deploying such powerful tools across multiple agencies. The involvement of the intelligence community in establishing safeguards suggests officials are taking a comprehensive approach to risk management.