Payward, the parent company of crypto exchange Kraken, has joined Anthropic's Project Glasswing, an initiative focused on AI-powered cybersecurity. The partnership positions Payward to leverage Anthropic's advanced AI models to identify and remediate security vulnerabilities across its infrastructure.
Under the arrangement, Payward will deploy Anthropic's cybersecurity-focused AI tools to conduct threat hunting and vulnerability assessments. The company plans to publish open-source findings from these security operations, contributing to broader industry knowledge around digital asset infrastructure protection.
Project Glasswing represents a broader effort by Anthropic to apply large language models toward defensive security work. By bringing Payward into the fold, the project gains a major player in the crypto sector whose security posture directly affects hundreds of thousands of users managing digital assets worth billions.
The timing reflects growing pressure on crypto custodians and exchanges to demonstrate robust security protocols. Kraken, one of the industry's largest regulated exchanges, handles significant trading volumes and custody. A security breach would carry reputational and financial consequences far beyond the immediate incident.
Payward's participation signals that established crypto firms now view AI-powered security scanning not as optional but as table stakes. The open-source component of the arrangement also suggests a willingness to share threat intelligence that benefits competitors and smaller players lacking comparable security resources.
Anthropic's involvement adds credibility to the cybersecurity approach. The AI safety-focused company has positioned itself as building more reliable and interpretable language models compared to competitors. Applying these models to security hunting translates that positioning into concrete defensive infrastructure.
The partnership sits at the intersection of two rapidly maturing sectors. Crypto exchanges face constant pressure from sophisticated attackers targeting custodial systems and smart contracts. Anthropic's Claude models bring computational horsepower and reasoning capabilities to detect attack patterns humans might miss.
For