Cohesity, the AI-powered data security company, has been granted access to Anthropic’s Claude Mythos Preview as a participant in Project Glasswing, Anthropic’s programme giving trusted organisations access to its most advanced frontier model to secure critical software infrastructure.

Cohesity will use Claude Mythos Preview to surface, validate and remediate potential vulnerabilities across its platform and product portfolio. The company serves more than 70 per cent of the Fortune Global 500 and some of the world’s largest governments, making it a high-value target for adversarial actors and a natural candidate for the programme.

AI accelerating both attack and defence

Project Glasswing is designed for organisations that build and operate critical software infrastructure, giving them restricted access to Anthropic’s most capable model to proactively harden their systems. Findings from across the partner community are shared to benefit the broader security ecosystem.

For Cohesity, the programme represents an extension of its security-first posture. By testing Claude Mythos Preview against its own systems, the company aims to shorten the window between when a vulnerability could exist and when it is remediated — a window that is narrowing as AI accelerates both the discovery and weaponisation of software flaws.

“Many of the largest organisations in the world, including more than 70% of the Fortune Global 500 and some of the largest governments, rely on Cohesity for their data resilience. That responsibility drives everything we do. Being part of Project Glasswing lets us put Mythos Preview to work directly on our platform, finding and fixing issues proactively, so we can maintain the resilience across our platform and product portfolio for the organisations that depend on us.” — Sanjay Poonen, CEO and President, Cohesity

Frontier models as security tools

The announcement reflects a broader shift in enterprise cybersecurity: frontier AI models are increasingly being deployed not just to accelerate productivity, but to stress-test the very infrastructure that enterprises and governments depend on. Cohesity’s participation signals that AI-powered data security platforms are now themselves becoming proving grounds for next-generation AI capabilities.

Cohesity’s platform is currently trusted by customers in more than 140 countries and is backed by investors and partners including NVIDIA, Amazon, Google, IBM, Cisco and HPE.