Project Glasswing and Anthropic's Quest to Build a Secure Internet

Anthropic's Mythos has unearthed several loopholes in existing software. Through Project Glasswing, the company is leading a crusade to secure internet in the AI era.

Cybersecurity in the AI Era: Project Glasswing has fundamentally transformed the traditional idea of cyber preparedness.

The revelations made by Mythos Preview, Anthropic’s powerful new model has upended the world of cybersecurity over the past few weeks. Mythos opened a Pandora’s box in April 2026 when it unearthed multiple loopholes in existing software and cybersecurity architectures. The unravelling of these loopholes, which would have remained unknown if it wasn’t for Mythos, have left security planners around the world in shock and awe. Fears are rife that an inevitable future explosion of similar AI models can offer myriad avenues to hackers and malicious actors to disrupt military and economic activity.

This fear is now catalysing an introspection into how cybersecurity should adapt to face new dangers in a threat matrix consisting of an infinite number of unknowns. Anthropic itself is guiding this introspection and has assumed a leading role in shaping a new cybersecurity paradigm for the AI era.  

The company is spearheading this crusade through the Project Glasswing, a new cybersecurity collaboration initiated in April 2026. Anthropic conceptualised the initiative as a unique inter-industry collective with a purpose – to use Mythos as a key to assess existing vulnerabilities in software. Having initially given access to 40 US-based companies at the time of its launch, Anthropic is now set to scale up Project Glasswing to include over 150 partners from 15 countries. In doing so, it is standing up a new kind of “private-led” mechanism within the existing global AI governance architecture.

Project Glasswing Revelations and Cybersecurity

Anthropic’s initial update on Project Glasswing published on 22 May 2026 underlines how Mythos has fundamentally changed the traditional cybersecurity paradigm. This is because AI is now better at quickly identifying vulnerabilities. The traditional understanding on cyber preparedness has itself been built around the process of identifying vulnerabilities. Moreover, within current protocols, there is a long time-lag between when a vulnerability is identified and a ‘patch’ is created to address it. The advent of models such as Mythos, “shrink the time and cost required to find and exploit vulnerabilities, magnifying the risk associated with these time lags”.

“A patch is simply an update to an Operating System or software which seeks to address flaws and vulnerabilities within a product.”

Consequently, cybersecurity planning is now all about addressing this time lag and responding in time to verify and patch any number of vulnerabilities. With the time lag persisting, there is a high risk that adversaries could exploit the loopholes unearthed by similar tools. In the long run, however, “Mythos-class models will enable developers to build far more secure software by catching bugs before they are deployed”.

Anthropic’s Quest to Create a Secure Internet

However, this cannot be done without a major collaborative effort to understand and address the overwhelming data on vulnerabilities coming to light. It is this sort of a collaborative effort that Anthropic seeks to lead from the helm of Project Glasswing.

Anthropic’s CEO Dario Amodei has touted Project Glasswing as “a real opportunity to create a fundamentally more secure internet and world than we had before the advent of AI-powered cyber capabilities”. In keeping with this pursuit is the decision to expand Glasswing beyond American soil. The expanded list of collaborators who are a part of Glasswing include companies operating in fields such as healthcare, water, power and communications. Moreover, the new partners shall only gain access to Mythos following a security clearance.

AI has fundamentally reshaped the limits of state capacity vis-à-vis private players. As the gatekeeper of the coveted tool which can both undermine and defend cybersecurity, Anthropic is already shaping the future narrative on cybersecurity, as much as any state government.

So far, no non-American entity – except for the UK’s AI Security Institute – had been allowed to access Mythos. The company had also offered the European Union access to Mythos, helping the bloc understand the risks emanating from similar models. However, this latest bid to take Glasswing global indicates Anthropic’s ambition to lead the global quest for a secure internet.

Anthropic is Rewriting the Rules of Responsive AI Collaboration

In spearheading this venture, Anthropic is re-writing the rules for how global cooperation on AI can maintain responsiveness with respect to the challenges thrown open by rapid technological advances. Government-led AI multilateralism, while alive, has not yet gained credibility with regards to its ability to quickly respond to threats from AI. Through Project Glasswing, Anthropic is seeking to achieve an important precedent not just for cybersecurity, but also for how AI collaboration and diplomacy can be rethought.

This situation with Mythos also goes to show how AI has fundamentally reshaped the limits of state capacity vis-à-vis private players. As the gatekeeper of the coveted tool which can both undermine and defend cybersecurity, Anthropic is already shaping the future narrative on cybersecurity, as much as any state government. Moreover, Project Glasswing shall serve as a central point of information for governments and powerful corporates recalibrating their cybersecurity strategies in the AI era.

Ultimately, a threat from technology can only be neutralised through an antidote from the same technology. Private actors are today in possession of the required technical know-how and the ability to innovate, in many cases having greater capacities than states. They hence have a central role in addressing threats from AI alongside governments. Let us all just hope that the Anthropic’s of the world truly understand that ‘with great power, comes great responsibility’.

Disclaimer: The article expresses the author’s views on the matter and do not reflect the opinions and beliefs of any institution they belong to or of Trivium Think Tank and the StraTechos website.

Anupama Vijayakumar

Anupama is the Director (Research) of Trivium Think Tank, Thiruvananthapuram, Kerala, India. She is also the Editor of the StraTechos website.