Anthropic Launches Project Glasswing to Defend Against AI Cyberthreats
Anthropic’s new powerful AI model Claude Mythos will be given to 12 companies to identify cybersecurity gaps.
WASHINGTON, April 9, 2026 – Anthropic’s announcement of Project Glasswing on Tuesday made one thing clear: AI is now as capable as any highly trained experts at identifying software vulnerabilities.
Project Glassjaw, a partnership between Anthropic, Apple, Nvidia, Amazon Web Services, J.P. Morgan Chase, Google and others will provide Anthropic’s new frontier model “Claude Mythos” to fix vulnerabilities or weaknesses in their foundational systems.
Anthropic’s relationship with the federal government remains soured after Wednesday’s decision by the D.C. Court of Appeals which allowed the Pentagon to continue its blacklist of the company after a dispute between the company and Secretary of Defense Pete Hegseth on the deployment of Claude in fully autonomous weapons.
“Mythos Preview has already found thousands of high-severity vulnerabilities, including some in every major operating system and web browser,” the announcement said. “Given the rate of AI progress, it will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely.”
In a post on LinkedIn, Rahul Patil, CTO of Antropic, noted that many companies have stated to him that speed in which code is being written today is outpacing how fast companies are able to check for vulnerabilities. Some of these vulnerabilities, found by Claude Mythos, are long standing according to the announcement.
In testing, the model found several vulnerabilities in the Linux kernel, the software that runs most servers in the world that would allow hackers to gain complete control over machines. In addition the model found a 27-year old vulnerability in OpenBSD, one of the most security hardened operating systems in the world that runs firewalls.
“By prioritizing defensive access to these powerful capabilities, Anthropic is helping us ensure that while intelligence is being weaponized, the defenders are the ones with the superior stack. AI becomes the defender.” said Nikesh Arora, CEO of Palo Alto Networks, a leading cybersecurity company.
Anthropic also noted that it was in conversation with the U.S. Government, and was prepared to assist state, and local officials with the use of Claude Mythos to secure critical infrastructure.
However, the release noted that the most important reason for the United States and its allies to be leaders in AI is to stay ahead of the malicious actors.
“Cyber defense at this scale is a team effort. Frontier labs, software companies, security researchers, open-source maintainers, and governments all working together is how defenders will stay ahead,” said Daniela Amodei, president of Anthropic.
Anthropic has promised to share what they learn through this project with the public.
Member discussion