Anthropic said it is expanding Project Glasswing to about 150 additional organizations across more than 15 countries, widening access to its AI-powered cyberdefense program beyond major tech companies and into critical infrastructure sectors.
The June 2 move expands access to Claude Mythos Preview beyond its original Big Tech and hyperscaler circle. The new cohort brings in sectors that were notably absent from the original launch: power, water, healthcare, communications, and hardware. Anthropic did not disclose the names of the new partner organizations or identify the countries involved
The AI company said a successful attack on most new partners could affect more than 100 million people. With this move, each partner gets gated access to Claude Mythos Preview, the company’s most capable model, which remains in controlled research preview because of its offensive cyber capabilities.
Analysis
What This Means for ERP Insiders
The threat model has changed permanently, for every enterprise, not just Glasswing partners. If Mythos-class models are six to twelve months away from general availability across multiple vendors, then every organization’s security posture is already obsolete against the threat they will face by end of 2026. Glasswing partners are being handed a head start, but the underlying message applies universally: the speed and sophistication of AI-assisted attacks will outpace any manual or legacy-automated defense. Enterprise CISOs who are not actively piloting AI-native security tooling today are building a gap they may not be able to close.
Why Patching Is Now the Bottleneck
The expansion also underscores a more immediate operational problem: organizations may soon be able to find serious flaws much faster than they can fix them.
Anthropic acknowledged that the bottleneck in the program is now disclosure, verification, and patching. The company said it is actively working with third parties to scale up the reviewing and patching of open-source software vulnerabilities and is developing shared standards for disclosing findings to maintainers in ways that are easier to act on.
“AI is accelerating both innovation and cyber risk. The challenge is no longer just finding vulnerabilities faster; it’s helping businesses prioritize, remediate and stay protected before risk reaches customers,” said Sachin Puri, CEO of Network Solutions, which provides domains, websites, hosting, email, and online identity services.
To give the broader market a path into AI-assisted security without full Mythos Preview access, Anthropic has also launched Claude Security, a product built on its latest publicly available frontier models, including Claude Opus 4.8, for codebase scanning and patch suggestions.
The company has also made available, on request to trusted security teams, the internal tooling its Glasswing partners use to accelerate vulnerability discovery.
Analysis
What This Means for ERP Insiders
The “defender’s advantage” framing is a strategic bet, not a guarantee, and the patch bottleneck is the real story. Project Glasswing has already surfaced a problem that undermines that framing: finding vulnerabilities at scale is now easy; fixing them is not. How Anthropic and its partners solve the disclosure-to-patch pipeline is the real test of whether Project Glasswing delivers on its promise, or inadvertently creates a sprawling vulnerability database that becomes a target in itself.
How Anthropic Is Packaging AI Cyberdefense
Anthropic’s longer-term stated goal is a “Cyber Verification Program” that would extend Mythos-class capabilities to a much wider pool of organizations for specific defensive tasks, a kind of credentialed access model that stops short of general release.
Access to Mythos Preview for Glasswing participants is priced at $25 per million input tokens and $125 per million output tokens, available through the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Anthropic has committed $100 million in model usage credits to cover the program.
The expansion announcement also landed alongside two other significant disclosures: Anthropic has filed a confidential draft S-1 with the SEC, signaling a path toward a public offering, and the company raised a $65 billion Series H round at a $965 billion post-money valuation. The combination of an IPO runway and a flagship security program with 200-plus institutional partners paints a picture of a company building toward enterprise dominance in a market that is reshaping around AI-native defense.
Analysis
What This Means for ERP Insiders
Glasswing is Anthropic’s most concrete enterprise moat, and its IPO story. The S-1 filing is not coincidental timing. Project Glasswing locks some of the world’s most critical institutional names, AWS, Microsoft, Google, JPMorganChase, Cisco, national infrastructure operators, into Anthropic’s ecosystem at the model level and at the tooling level. These are multi-year, operationally embedded relationships, not API subscriptions. For a company heading toward a public offering at a near-trillion-dollar valuation, Glasswing provides something venture-backed AI labs rarely have: a defensible, mission-critical reason for the world’s largest enterprises to stay. Investors watching the S-1 should pay as much attention to Glasswing’s partner retention and patching metrics as to model benchmarks.
