TL;DR:
- Payward formally joined Project Glasswing, the Anthropic security initiative launched in April 2026.
- The company will integrate the defensive artificial intelligence model Claude Mythos 5 across all its operational environments in the coming weeks.
- The program includes the responsible disclosure to external developers of any vulnerability detected in open-source software.
This Monday, Payward, parent company of the exchange Kraken, confirmed it is joining Anthropic’s Project Glasswing program to deploy Claude Mythos 5 across its defensive cybersecurity operations.
The initiative grants the firm access to advanced artificial intelligence capabilities to audit code and detect critical vulnerabilities. Kraken’s parent company stated that the model will be progressively implemented across its platforms over the coming weeks to support the efforts of its technical teams.
Technical Integration and Risk Mitigation in Financial Infrastructure
According to Payward’s official statement, access to this system follows a United States regulatory authorization allowing the deployment of the model in organizations managing critical infrastructure. The firm operates the technological backbone behind brands like Kraken, CF Benchmarks, and NinjaTrader, managing shared settlement, risk, and regulatory compliance systems.
The deployment of Claude Mythos 5 will build upon the entity’s existing security protocols, which hold SOC 2 and ISO 27001 certifications.
Within the firm’s operational framework, data generated by the model will directly feed into the efforts of its internal red team and blue team divisions. According to Anthropic’s technical reports, members of Project Glasswing have identified thousands of high or critical severity security flaws since the program’s inception in April 2026.
Institutional documentation highlights that the model will evaluate third-party libraries and open-source repositories integrated into its architecture. Under Payward’s responsible disclosure protocol, flaws detected in open-source dependencies will be reported directly to each project’s maintainers to facilitate global patching.
With this move, Payward joins a select group of organizations chosen by Anthropic across technology and banking sectors, featuring entities such as Amazon Web Services, Microsoft, and JPMorganChase.
Program guidelines establish that model-assisted code reviews will begin immediately across the company’s production environments and transactional modules.
