Project Glasswing: Securing critical software for the AI era
anthropic apple open-source
| Source: Mastodon | Original article
Anthropic has unveiled Project Glasswing, a collaborative effort to harden the world’s most critical software against AI‑driven attacks. The initiative brings together Apple, Google, Microsoft, Amazon and a dozen other tech heavyweights under a shared defensive framework built on Anthropic’s unreleased Mythos model. Within weeks of its launch, the consortium has already identified and patched thousands of high‑severity vulnerabilities that could be weaponised by sophisticated threat actors using large language models to automate exploit discovery.
The move marks the first coordinated, industry‑wide response to the emerging risk that generative AI poses to the software supply chain. By feeding a powerful Claude‑style model with codebases from participating firms, Glasswing can scan, fuzz and propose mitigations at a scale no human team could match. Anthropic’s statement frames the project as “the AI era’s Manhattan Project for software security,” echoing the urgency expressed in our April 8 coverage of instant‑software threats. The partnership also signals a shift from competitive secrecy to collective resilience, acknowledging that no single company can defend the global digital infrastructure alone.
What follows will test the initiative’s governance and impact. Observers will watch how the shared vulnerability database is managed, whether participation expands beyond the founding members, and how regulators respond to a quasi‑public‑private security consortium. The effectiveness of Mythos‑driven remediation will be gauged against real‑world incidents, especially as nation‑state actors accelerate AI‑assisted cyber campaigns. Finally, the industry will be keen to see if Glasswing’s model can be open‑sourced or replicated, setting a new standard for collaborative cyber‑defence in an AI‑saturated landscape.
Sources
Back to AIPULSEN