Anthropic Launches Project Glasswing With Claude Mythos Preview

What happened

Anthropic announced Project Glasswing on X and published full details on April 7, 2026. The company describes it as an initiative to secure the world’s most critical software for the AI era, built around Claude Mythos Preview. Anthropic named Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks among its launch partners, and said it had extended access to more than 40 additional organizations that build or maintain critical software infrastructure.

Anthropic says participants will use Claude Mythos Preview in defensive security work and that the company will share what it learns with the broader industry. It also committed up to $100 million in usage credits and $4 million in donations to open-source security organizations. According to Anthropic, Mythos Preview is already identifying large numbers of zero-day vulnerabilities across critical infrastructure, though that claim should still be understood as company-reported performance.

Why it matters

The deeper significance of the launch is strategic. Anthropic is explicitly treating frontier model deployment as a cybersecurity timing problem: if strong capability is coming either way, defenders need earlier access, better tooling, and coordinated evaluation before offensive use catches up. That is a more concrete posture than generic statements about AI safety.

• The partner list spans cloud providers, security vendors, semiconductor companies, financial institutions, and open-source infrastructure stewards.
• Anthropic says the model will be available through the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
• The initiative combines a gated research preview, ecosystem funding, and public commitments to share lessons learned.

That cross-industry structure matters because software security is not confined to one vendor or one stack. Critical systems depend on a supply chain of clouds, chips, libraries, open-source maintainers, and enterprise operators. By framing Glasswing around that shared surface area, Anthropic is making the case that AI-assisted defense has to scale across the ecosystem, not just within one internal security team.

The obvious open questions remain. External validation matters: can Mythos Preview sustain its claimed advantage in independent settings, and can a defense-first rollout actually reduce real-world exploitation windows? Even with those questions, Glasswing is a notable sign that frontier labs are moving from abstract cyber claims toward structured defensive programs with named partners, budgets, and deployment channels. Original source: Anthropic.