Anthropic Expands Access to Claude Mythos, Adding 150 Critical Infrastructure Organizations to Project Glasswing

Anthropic announced on Tuesday that it is expanding access to Claude Mythos Preview, its most capable and most restricted AI model, through Project Glasswing. Roughly 150 new organizations across critical infrastructure sectors will gain access to the model, which is designed to help defend against sophisticated cyber threats.

The company laid down sobering words in its announcement: "A successful attack on their codebase could be catastrophic." For most partners, the stakes are existential, and Anthropic is betting that its most advanced model can help prevent such outcomes.

Project Glasswing and Claude Mythos Preview

Project Glasswing is Anthropic's initiative to provide select organizations with early access to Claude Mythos Preview, a model that has been tightly restricted since its development. The expansion adds roughly 150 new organizations from sectors including energy, finance, healthcare, and transportation. These organizations will use the model to analyze code, detect vulnerabilities, and respond to threats in real time.

• Claude Mythos Preview is Anthropic's most capable model for cybersecurity tasks, but it is also the most restricted due to safety concerns.
• The European Union Agency for Cybersecurity (ENISA) is among the new partners, a result of "strong bilateral cooperation" between the European Commission and Anthropic, according to Dark Reading.
• Anthropic has not disclosed the full list of new partners, but the expansion targets critical infrastructure sectors that are frequent targets of state-sponsored and criminal cyberattacks.
• The model is designed to assist with code review, threat hunting, and incident response, reducing the time needed to identify and patch vulnerabilities.

Implications for Critical Infrastructure Security

The expansion comes as critical infrastructure operators face increasing pressure from ransomware groups and nation-state actors. By giving these organizations access to Claude Mythos Preview, Anthropic aims to close the gap between attacker capabilities and defender resources. However, the company also acknowledges the risks: a compromised partner could expose the model to adversaries, potentially leading to catastrophic outcomes.

Anthropic has not set a timeline for broader availability of Claude Mythos Preview. The company says it will continue to evaluate partners on a case-by-case basis, prioritizing organizations with strong security postures and clear use cases. For now, the focus remains on critical infrastructure, where the impact of a successful attack could ripple across economies and societies.