Anthropic promotes Project Glasswing to strengthen global software cybersecurity with the Claude Mythos model

Anthropic has presented the new Proyecto Glasswing, a cybersecurity initiative that brings together technology companies such as Amazon Web Serivces, Apple, Google, Microsoft and Nvidia, to protect “the world’s most critical software”, based on your model Claude Mythoscapable of identifying high severity vulnerabilities in all major operating systems and web browsers.

He ‘software’ most used on a daily basis, responsible for relevant issues such as banking systems, the storage of medical records or the interconnection of logistics networks, may contain failures that, if discovered, can allow cybercriminals to hijack systems, disrupt operations, or steal datawith serious consequences for end users.

In this framework, the artificial intelligence (AI) company has highlighted how, currently, AI models he achieved a level of programming ability that surpasses most users in the detection and exploitation of software vulnerabilities.

That is, they have high capabilities to find these flaws and look for ways to exploit them effectivelydamaging both large infrastructure and government agencies, as well as smaller-scale attacks targeting hospitals or schools.

To face this scenario, Anthropic has trained a new cutting-edge model called Claude Mythos Preview which, although not yet published, has already detected thousands of high-severity vulnerabilities in the main operating systems and web browsers.

As the technology company specified in a statement, this model focused on identifying vulnerabilities “could transform cybersecurity” and, under this premise, it has Project Glasswing announced

This is a new initiative that brings together relevant companies in the sector technology such as Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia and Palo Alto Networks, so that they can have this model and join together in a effort to protect the “world’s most critical software.”

“Given the pace of advancement of AI, it will not be long before these capabilities spread, possibly beyond the actors committed to their safe implementation,” Anthropic has warned, while explaining that Project Glasswing is “a urgent attempt to use these capabilities for defensive purposes.”

CLAUDE MYTHOS TO DETECT VULNERABILITIES

Specifically, the model Claude Mythos Preview has already been used in various tests and the results show that there are some vulnerabilities that have survived “decades of human review and millions of automated security tests.

This is because many software flaws go unnoticed for years because their detection and exploitation required specialized knowledge. However, with next-generation AI models, the level of expertise required to find such vulnerabilities has been drastically reduced.

Specifically, Anthropic has used Claude Mythos to identify thousands of zero-day vulnerabilities, many of them critical, in various operating systems, web browsers and relevant programs.

For example, Mythos Preview discovered a vulnerability that had been in OpenBSD for 27 years, one of the operating systems considered one of the most secure in the world and used to run firewalls and other critical infrastructure. This vulnerability allowed any machine running the system to be remotely blocked.

Another example is a 16 year old vulnerability old found in the video encoding and decoding program FFMpegas well as several Vulnerabilities in the Linux kernel.

These vulnerabilities were identified by the Mythos Preview totally autonomouswithout the need for any human intervention, as the company has stressed, and they have already been corrected.

GLASSWING PROJECT FOR DEFENSIVE SECURITY

With all this, the Glasswing Project will allow the founding partners to use Mythos Preview in their defensive security tasks, sharing their learnings with each other so that “the whole industry benefits.”

Additionally, Anthropic has expanded access to this model to more than 40 organizations responsible for developing or maintaining critical software infrastructuresso that they have the capabilities to analyze and protect both their own and open source systems.

The technology company has also announced that will allocate up to $100 million in Mythos Preview usage credits for these initiativesas well as an additional $4 million in direct donations to open source security organizations.

“The Glasswing project represents an important step in giving defenders a lasting advantage in the next era of AI-driven cybersecurity,” Anthropic said, while clarifying that it is “a starting point” and that AI developers, software companies, security researchers and governments around the world “have a fundamental role to play.”