The New York Stock Exchange has deployed Claude Mythos Preview, an Anthropic model offered through the broader initiative known as Project Glasswing, to identify and patch cybersecurity vulnerabilities across its systems.

NYSE President Lynn Martin testified before the House Financial Services Committee on September 2, 2026, explaining how the technology enabled the exchange to rapidly detect and remediate security flaws. According to her testimony, some of those vulnerabilities had been present in the codebase for decades — a common problem for financial market infrastructure, where exchanges and clearing systems run on long-lived codebases that are difficult to audit manually at scale. The testimony also placed the deployment in a regulatory context: exchanges like NYSE are designated systemically important financial market utilities, and congressional committees have increased scrutiny of their operational resilience and cybersecurity posture in recent years.

What Project Glasswing actually does

Anthropic launched Project Glasswing on April 7, 2026, granting selected organizations access to Claude Mythos Preview, a model purpose-built for deep code analysis and vulnerability detection. The program began with heavyweight partners including AWS, Apple, Google, and Microsoft before expanding to a wider group.

Intercontinental Exchange, NYSE's parent company, formally joined the initiative on June 3, 2026, and deployed Claude Mythos Preview across all of its platforms.

Partners testing the model reported that bug discovery rates improved by more than tenfold compared with previous methods. Thousands of critical vulnerabilities were surfaced, some of which had gone undetected for years.

Claude Mythos Preview is priced at $25 per million input tokens and $125 per million output tokens. Anthropic has also committed up to $100 million in usage credits and $4 million in security donations to support participating organizations.

From Wall Street to hospitals and power grids

By early June 2026, Project Glasswing had expanded to include roughly 150 additional organizations spanning the healthcare, power utility, and communications sectors — industries the U.S. government designates as critical infrastructure, where software flaws carry potential consequences beyond any single company. The breadth of adoption suggests the program will be a test case for whether AI-assisted code auditing becomes a standard practice for securing legacy software across the economy, and further congressional testimony or partner disclosures may indicate how widely the approach spreads.