Anthropic Launches Mythos AI in India, Expands Cybersecurity Initiative Globally

Anthropic has significantly expanded its high-security AI initiative, Project Glasswing, bringing its most advanced model, Claude Mythos Preview, to India and several other countries to strengthen global cybersecurity infrastructure.

Anthropic Broadens Global Reach of Project Glasswing

Anthropic, a leading artificial intelligence company headquartered in San Francisco, has announced a major expansion of its exclusive cybersecurity programme, Project Glasswing. The initiative, initially launched in April with a limited group of around 50 partners, has now grown to include approximately 150 organisations worldwide. This marks the largest expansion of the programme to date, reflecting the growing urgency around digital security and AI-driven threat detection.

India is among the newly added countries gaining access to this programme, alongside major economies such as Canada, Australia, France, Germany, Japan, and South Korea. The expansion spans over 15 nations, bringing together organisations that manage critical infrastructure and digital systems used by millions globally.

Strategic Inclusion of Critical Global Organisations

The newly expanded cohort includes a diverse set of high-impact organisations across industries. According to reports, participants now include major technology firms like Samsung, SK Telecom, and SK Hynix, cybersecurity company Okta, financial network Swift, and settlement platform Euroclear. Additionally, Intercontinental Exchange, which operates the New York Stock Exchange, has also joined the programme.

Government and defence-related bodies such as NATO and the European Union Agency for Cybersecurity (ENISA) are also part of the initiative. These organisations play crucial roles in maintaining national and international security, making their inclusion particularly significant.

Anthropic has emphasised that many of these participants are vendors whose technologies form the backbone of services used by governments, enterprises, and millions of end users. This interconnectedness means that vulnerabilities within these systems can have far-reaching consequences.

Focus on Strengthening Critical Infrastructure Sectors

A key aspect of this expansion is the inclusion of industries that were previously underrepresented. These sectors include energy, water supply, healthcare, telecommunications, and hardware manufacturing—all of which are essential to national critical infrastructure.

Anthropic highlighted that a successful cyberattack on these systems could potentially impact more than 100 million people. This underscores the importance of proactive vulnerability detection and reinforces the need for advanced AI tools to safeguard these sectors.

What Makes Claude Mythos Preview Unique?

At the heart of Project Glasswing lies Claude Mythos Preview, Anthropic’s most advanced and tightly controlled AI model. Unlike publicly available AI systems, this model is specifically designed for deep cybersecurity analysis.

Claude Mythos Preview is capable of scanning vast and complex codebases to identify hidden vulnerabilities—some of which may have remained undetected for decades. Its advanced reasoning and coding capabilities enable it not only to detect flaws but also to simulate potential exploits.

This dual capability is precisely why Anthropic has chosen not to release the model to the public. While it can significantly enhance defensive cybersecurity efforts, it also possesses the potential to be misused if it falls into the wrong hands.

Early Results Show Massive Impact

The initial phase of Project Glasswing has already delivered impressive results. In just a few weeks, the first group of approximately 50 partners collectively identified over 10,000 high and critical security vulnerabilities.

One notable example is Cloudflare, which discovered around 2,000 bugs within its systems, including 400 classified as high or critical severity. These findings highlight the model’s effectiveness in identifying complex security issues that traditional tools may overlook.

Additionally, Claude Mythos Preview demonstrated its advanced capabilities by successfully completing both cyber range simulations conducted by the UK’s AI Security Institute, marking a significant milestone in AI-driven cybersecurity testing.

Why the Expansion Is Urgent

Anthropic has acknowledged that the rapid expansion of Project Glasswing is driven by a sense of urgency. The company anticipates that within the next six to twelve months, other AI developers may create models with similar capabilities.

However, there is concern that such models could be released without adequate safeguards, potentially enabling malicious actors to exploit vulnerabilities at scale. This creates a pressing need to strengthen defensive systems before offensive capabilities become widely accessible.

Anthropic’s approach aims to stay ahead of this curve by equipping trusted organisations with advanced tools to identify and fix vulnerabilities before they can be exploited.

A Dual Mission: Detection and Remediation

The company has outlined a long-term vision for Project Glasswing that goes beyond simply identifying vulnerabilities. Anthropic plans to gradually shift the programme’s focus toward supporting the full lifecycle of cybersecurity management.

This includes responsible disclosure of vulnerabilities, faster patch development, and efficient deployment of fixes across systems. By doing so, the initiative aims to create a more resilient and proactive cybersecurity ecosystem.

Introduction of Claude Security for Enterprises

Alongside the expansion of Project Glasswing, Anthropic has also launched a new product called Claude Security. Unlike Mythos Preview, this solution is built on publicly available models and is designed for enterprise customers.

Claude Security offers organisations advanced tools for identifying and managing security risks within their systems. Additionally, Anthropic is making internal scanning tools available to qualified security teams upon request, further enhancing accessibility to AI-powered cybersecurity solutions.

Shaping the Future of AI-Driven Cybersecurity

Anthropic’s latest move signals a broader shift in how artificial intelligence is being deployed in cybersecurity. By restricting access to its most powerful model while selectively expanding its reach, the company is attempting to balance innovation with responsibility.

As cyber threats become more sophisticated, initiatives like Project Glasswing could play a crucial role in protecting global digital infrastructure. The expansion to India and other nations marks an important step toward building a more secure and resilient digital future.