News reports published by Reuters stated that Microsoft announced a major strategic move on Wednesday to integrate the preview version of Anthropic’s “Claude Mythos” model into its security development lifecycle framework. This step represents an official adoption of this highly powerful and advanced intelligent model in detecting security vulnerabilities across all of the company’s enterprise and major corporate software.
This strategic initiative makes Microsoft one of the first major technology companies to directly integrate this advanced model into its internal software engineering operations. Anthropic has described this model as “vastly superior to any other artificial intelligence model in cyber capabilities at the present time.” The integration of the “Claude Mythos” model into the security development lifecycle framework is expected to help rapidly identify weaknesses and develop effective fixes for them at very early stages of the software development cycle, reducing risks before products are launched on the market.
An ultra-capable model shrouded in strict confidentiality
This technological integration comes as part of a closed and restricted initiative known as “Project Glass Wing,” which was announced on April 7. The initiative aims to secure critical software infrastructure worldwide and brings together under its umbrella an elite group of the largest and most important technological and financial entities, including:
- Amazon Web Services
- Apple
- Broadcom
- Cisco
- CrowdStrike
- JPMorgan Chase
- Linux Foundation
- Microsoft
- Nvidia
- Palo Alto Networks
Anthropic stated that it has absolutely no plans to make the preview version of the “Claude Mythos” model available to the public or for general use, citing severe risks that its capabilities could be weaponized by malicious actors and cyber attackers. In a draft blog post accidentally published last March, the company clearly warned that this model “heralds an upcoming wave of artificial intelligence models that can exploit security vulnerabilities in ways that far exceed the efforts of defenders and cybersecurity professionals.”>
Discovery of historical vulnerabilities and autonomous cyberattacks
During testing phases prior to the launch, the model demonstrated terrifying capabilities, managing to identify thousands of zero-day vulnerabilities—flaws that were previously unknown to developers. According to Anthropic data, these vulnerabilities spanned all major operating systems and web browsers available today, and the oldest vulnerability the model managed to discover had existed in the OpenBSD system for about 27 years without being noticed.
In this context, Logan Graham, who leads offensive cyberspace research at Anthropic, told NBC News that the model not only discovers hidden software flaws, but is advanced enough to connect a chain of vulnerabilities and exploits completely independently to breach complex systems without human intervention, making it an exceptional tool in network security assessment.
A growing strategic partnership between Microsoft and Anthropic
Microsoft has been significantly deepening its strategic partnership with Anthropic over recent months. In early 2026, the company made Anthropic models available as a default option across the advanced Microsoft 365 Copilot suite and other products, with Anthropic joining as an approved sub-processor within Microsoft’s enterprise data framework. Alongside this, Microsoft poured a massive $5 billion investment into Anthropic, securing in return a $30-billion commitment to utilize cloud computing capabilities from its Azure platform.
Developing the security development lifecycle framework
This integration extends to touch the core of Microsoft’s engineering practices. The security development lifecycle framework, originally established as part of the Trustworthy Computing initiative in 2002, is a mandatory framework governing how all company software is designed, built, and tested. Last February, Microsoft published ambitious plans to expand this framework to address artificial intelligence security risks alongside traditional software challenges, affirming the company’s commitment to staying at the forefront of security innovation.
Financial support and operational costs for the new model
To support this unprecedented initiative, Anthropic has allocated $100 million in model usage credits to cover costs for participants in “Project Glass Wing.” Following the research preview period, participants will be able to access the preview version of “Claude Mythos” at a cost of $25 per million input tokens and $125 per million output tokens via the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and the Microsoft Foundry platform.
Frequently Asked Questions
Why did Microsoft decide to integrate the “Claude Mythos” model into its operations?
The company decided to integrate this advanced model to help identify security vulnerabilities and weaknesses in its software and develop fixes for them faster and at earlier stages of the software engineering cycle, radically raising the level of security.
Will the “Claude Mythos” model be available for public use?
No, Anthropic confirmed it has no intention of making the model available to the public due to its immense capabilities, which could pose a significant risk if they fall into the hands of malicious entities that might use them as cyber weapons to breach systems.
What is “Project Glass Wing” and who are its most prominent participants?
It is a closed and restricted initiative aimed at protecting critical software infrastructure, including major tech companies such as Microsoft, Apple, Amazon Web Services, Google, Nvidia, and the Linux Foundation, alongside financial institutions like JPMorgan Chase.
What is the cost of using the model for project participants?
After the research preview phase ends, the cost will be $25 per million tokens for input operations and $125 per million tokens for output operations.
