The point: Anthropic is expanding its AI-powered code security program to 150 new partners from critical infrastructure sectors, as the initial 50 partners have already identified over 10,000 critical vulnerabilities.
Anthropic is scaling up its Project Glasswing program to secure critical software. Following the discovery of more than ten thousand critical vulnerabilities by the initial 50 partners, approximately 150 new organizations are joining, including providers from energy, water, healthcare, and communications infrastructure.
Project Glasswing is an Anthropic initiative for collaboration on securing critical software globally. In April 2025, approximately 50 initial partners had access to Claude Myths Preview and used the model for codebase analysis. These partners identified more than ten thousand vulnerabilities with high or critical severity.
The expansion now includes approximately 150 additional organizations from more than 15 countries that must meet information security requirements. The new cohort covers industries that were underrepresented in the initial group: electric power, water management, healthcare, communications, and hardware. Many of the new partners are providers of code systems that organizations worldwide—including governments—depend on. A successful cyberattack on such codebases could affect over 100 million people for most partners and endanger national and global security.
Anthropic sees the need for this expansion in the fact that powerful AI models with cyber capabilities will be widely available in the foreseeable future. The company expects that within 6 to 12 months, many AI providers will develop Myths-like models and potentially release them without safeguards against misuse. Cyberattacks could then become more frequent and occur in less predictable forms.
In parallel with the partnership expansion, Anthropic is releasing certain tools for trusted security teams and a new product called Claude Security, which is based on publicly available models such as Claude Opus 4.8 and automatically scans codebases for vulnerabilities while suggesting patches. Long-term, Anthropic wants to support the industry in developing new standards and infrastructure—with the goal of shifting focus from mere vulnerability discovery to disclosure, remediation, and deployment of patched software.
Source: www.anthropic.com · Published June 1, 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.2.9.