Goldman Sachs is reportedly enhancing its cybersecurity measures by utilizing Claude Mythos in collaboration with Anthropic, according to Bloomberg. This initiative is part of Project Glasswing, which provides a select group of organizations, including Goldman Sachs and JPMorgan, with limited access to Claude Mythos for bolstering their defensive cybersecurity efforts. The model’s sophisticated reasoning capabilities enable it to identify critical vulnerabilities in major operating systems and browsers, including long-standing bugs, making it a valuable tool for mitigating cyber risks in the financial sector.
Anthropic: Anthropic develops advanced AI models focused on safety and reliability, including the Claude series. The company is collaborating with Goldman Sachs on deploying Claude Mythos for cyber defense enhancements. Through Project Glasswing, Anthropic provides restricted access to select organizations like banks and tech firms to secure critical systems proactively.
Claude Mythos: Claude Mythos is Anthropic’s unreleased frontier AI model preview featuring advanced agentic coding and reasoning skills that enable it to identify long-undetected vulnerabilities without specific cybersecurity training. It is being employed by Goldman Sachs to strengthen cyber defenses as part of limited-access initiatives. The model has demonstrated capabilities in autonomously chaining vulnerabilities and solving complex network attack simulations.
Goldman Sachs: Goldman Sachs is a leading global investment bank providing services in investment banking, securities, asset management, and consumer banking. The firm is reportedly ramping up its cyber defenses by utilizing Anthropic’s Claude Mythos AI model. This effort aligns with US authorities urging Wall Street banks, including Goldman Sachs, to test the model for enhanced security probing under Project Glasswing.
Project Glasswing: Anthropic’s initiative grants limited access to Claude Mythos Preview to organizations like Goldman Sachs, JPMorgan, and tech firms for defensive cybersecurity work.
Model Capabilities: Claude Mythos identifies critical vulnerabilities in major operating systems and browsers, including decades-old bugs, due to its general reasoning advancements.
Government Involvement: US Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened Wall Street leaders to encourage testing Anthropic’s Mythos for cyber risk mitigation.