Anthropic has launched Project Glasswing, a new cybersecurity effort built around its unreleased frontier model, Claude Mythos Preview. Launched on April 7, 2026, the initiative is aimed at securing ‘the world’s most critical software.’ The launch partners of the initiative include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.
The announcement comes as AI tools become stronger at finding software flaws. According to Anthropic, AI models have now reached a point where they can outperform most humans at finding and exploiting vulnerabilities. On its Project Glasswing page, the company said Claude Mythos Preview has already found thousands of high-severity flaws.
How Project Glasswing will Change Cybersecurity?
Individuals nowadays commonly use software containing their banking, medical, and other confidential information. However, minor or major flaws have always been threats to the confidentiality of such information. As a result, attackers could easily hijack systems, steal data, and interrupt operations.
As per Anthropic, Project Glasswing will focus on local vulnerability detection, black-box testing, endpoint security, and penetration testing. The partners in the program will use Mythos Preview to scan and secure first-party and open-source systems. The firm also extended access to more than 40 additional organizations that maintain critical software infrastructure. The model will be able to address the following components in the upcoming days-
- Vulnerability disclosure processes
- Software update processes
- Open-source and supply-chain security
- Software development lifecycle and secure-by-design practices
- Standards for regulated industries
- Triage scaling and automation
- Patching automation
Understanding Claude Mythos Preview
Claude Mythos Preview is a segment of Project Glasswing’s defensive security initiative. It has already detected thousands of high-severity vulnerabilities that have impacted major operating systems and web browsers. While identifying vulnerabilities, Mythos Preview has detected some incidents that have been actively prevailing for decades and survived human reviews and security tests.
Anthropic has used Mythos Preview for weeks now to detect several zero-day vulnerabilities, including the critical ones. Among them, the 27-year-old vulnerability in OpenBSD and the 16-year-old vulnerability in FFmpeg are significant.
Benchmarks Achieved by Mythos Preview:
Mythos Preview has shown strong results on several security and coding benchmarks. On CyberGym, it scored 83.1 percent, compared with 66.6 percent for Claude Opus 4.6. On SWE-bench Verified, it scored 93.9 percent, versus 80.8 percent for Opus 4.6.
The model also outperformed Opus 4.6 on Terminal-Bench 2.0, BrowseComp, OSWorld-Verified, and multiple other tests. Anthropic said the model’s gains come from stronger agentic coding and reasoning.
Availability and Accessibility of Mythos Preview:
As Anthropic suggests, the Claude Mythos Preview will not be made generally available. Access is limited to Project Glasswing partners and selected organizations. The company said participants will use the model through the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
Anthropic is also committing up to $100 million in usage of credits and $4 million in donations to open-source security groups. After the research preview, the model will be offered to participants at set usage prices of $ 25 or $125 per million input/output tokens.
Project Glasswing highlights the next phase of cyber risk. Anthropic highlights that the window between vulnerability discovery and exploitation is shrinking fast. Such activities make AI-assisted defense more urgent.
The firm said it will publish a public update within 90 days, along with any vulnerabilities and improvements it can disclose. Anthropic is also in talks about the model’s cyber capabilities with US government officials.
KnowledgeNile is a leading publication of tech updates and insights. Explore our news stories and stay updated in the technological age.
Also Read:
Anthropic’s Claude AI Now Generates Charts and Diagrams: What You Need to Know
