Anthropic has significantly expanded Project Glasswing, its cybersecurity initiative designed to help organizations identify and address critical software vulnerabilities using advanced AI models. The company is adding approximately 150 new organizations to the program, bringing total participation to roughly 200 partners.
The expansion follows Anthropic’s initial rollout in April, when around 50 organizations received access to Claude Mythos Preview, the company’s advanced cybersecurity-focused AI model. Since then, participants have used the system to scan software codebases and identify vulnerabilities. Anthropic recently reported that partners had already discovered more than 10,000 high- and critical-severity security flaws through the initiative.
The newly added organizations span more than 15 countries and include sectors such as energy, water, healthcare, communications, and hardware infrastructure. Many participants maintain software and systems that support governments, businesses, and critical infrastructure providers around the world.
Anthropic said organizations admitted to the program must meet specific security requirements before receiving access. According to the company, most partners operate software whose compromise could affect more than 100 million people, creating potentially significant national security and public safety consequences.
The initiative reflects Anthropic’s broader effort to ensure advanced AI capabilities provide an advantage to cybersecurity defenders rather than attackers. The company argues that increasingly powerful cyber-capable AI models are likely to become widely available across the industry within the next year, making it critical for organizations to strengthen defenses before those capabilities become commonplace.
Alongside the expansion, Anthropic announced plans to make additional security tools available to trusted cybersecurity teams and introduced Claude Security, a product that uses public Claude models to scan codebases and recommend software patches.
Preparing for the Next Cybersecurity Era
Anthropic believes cybersecurity is approaching a major transition as advanced AI models become capable of discovering vulnerabilities at unprecedented speed and scale. The company has repeatedly warned that future AI systems could dramatically increase both defensive and offensive cyber capabilities.
According to Anthropic, the primary challenge is no longer finding vulnerabilities but verifying, disclosing, and fixing them quickly enough. Many Project Glasswing participants are already using Mythos Preview not only to identify flaws but also to generate patches, conduct penetration testing, automate threat detection, and improve software security before deployment.
The company hopes these practices will help establish new operational standards for organizations preparing for an environment where highly capable AI-powered cyber tools become widely accessible.
Building a Defensive AI Ecosystem
Project Glasswing has evolved beyond a vulnerability discovery program into a broader effort to reshape how the cybersecurity industry responds to AI-driven threats. Anthropic said it is working with security researchers, software maintainers, governments, and infrastructure providers to develop processes for large-scale vulnerability management and coordinated disclosure.
Just recently, the company launched Claude Opus 4.8 and reinforced its position as one of the industry’s leading AI developers by raising a $65 billion Series H funding round at a $965 billion valuation, surpassing OpenAI as Silicon Valley’s most valuable AI company.
Looking ahead, Anthropic plans to continue expanding Project Glasswing, prioritizing critical infrastructure operators, maintainers of widely used open-source software, and additional security testing organizations. The company said its long-term goal is to ensure that increasingly powerful AI systems create a lasting advantage for defenders as cybersecurity enters a new phase shaped by frontier AI models.
