Anthropic is poised to launch an innovative AI model as part of a significant cybersecurity collaboration, dubbed Project Glasswing. This initiative brings together industry giants such as Nvidia, Google, Amazon Web Services, Apple, and Microsoft, among others. Project Glasswing aims to empower large corporations, and potentially government entities, to identify system vulnerabilities with minimal human intervention.
As part of this endeavor, Anthropic is granting its initial partners access to Claude Mythos Preview. This new general-purpose model is not slated for public release due to inherent security considerations. Newton Cheng, the cyber lead for Anthropic’s frontier red team, conveyed to The Verge that the model is designed to provide cyber defenders a crucial “head start” against potential adversaries. Partners will leverage the model to scrutinize their systems, pinpointing high-stakes vulnerabilities and aiding in their remediation. Access to the model is strictly controlled to prevent malicious actors from exploiting its capabilities to uncover weaknesses and launch attacks.
While Claude Mythos Preview was not specifically developed for cybersecurity applications, Anthropic stated in a release that the model's "strong agentic coding and reasoning skills" are the driving force behind its advanced cybersecurity capabilities. In an interview with The Verge, Newton Cheng, while declining to disclose specific details of the model's cybersecurity triumphs, referenced an Anthropic blog post revealing that Mythos Preview has recently identified "thousands of high-severity vulnerabilities, including some in every major operating system and web browser." The blog post further emphasized the model's autonomy, noting it identified vulnerabilities "and develop[ed] many related exploits — entirely autonomously, without any human steering," without mentioning human oversight in its cybersecurity assessments.
The existence of Claude Mythos Preview initially came to light last month through a data leak, an incident Anthropic attributes to human error. Dianne Penn, a head of product management at Anthropic, informed The Verge that the company is "taking steps in terms of solidifying our processes…" and clarified that the leak "was not related to software vulnerabilities in any way."
Mythos Preview will be exclusively available to Anthropic's Glasswing partners, a group that also includes major players like JPMorgan Chase, Broadcom, Cisco, CrowdStrike, the Linux Foundation, and Palo Alto Networks, alongside approximately 40 other organizations specializing in software infrastructure. Anthropic plans to subsidize the initial cost of its use, committing up to $100 million in usage credits. Additionally, Cheng announced direct donations of $4 million to both the Linux Foundation and the Apache Software Foundation. In the long term, as pressure mounts on AI companies to achieve profitability, this program could evolve into a paid service, generating a new revenue stream, provided its efficacy encourages continued adoption by companies.
Despite its recent high-profile disagreement with the Trump administration, Anthropic’s release also confirmed "ongoing discussions with US government officials about Claude Mythos Preview and its offensive and defensive cyber capabilities." When questioned by The Verge, Penn affirmed that the company had "briefed senior officials in the US government about Mythos and what it can do," reiterating Anthropic's "committed to working closely with all different levels of government." Cheng, while confirming Anthropic's engagement with the government, refrained from specifying the exact officials who received these briefings.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.