Payward, which is Kraken’s parent firm, has partnered with Anthropic’s Project Glasswing since Aug. 17 in cybersecurity activities. This collaboration enables Payward to have limited access to Claude Mythos 5, which is a cybersecurity AI model from Anthropic. In the coming weeks, Payward will be conducting a scan for vulnerabilities in its software environment. After reviewing the results, the security experts in the firm will sanction the implementation of any necessary corrections. The firm has not been mentioned by the firm whether Mythos 5 will scan production, isolated code, or test environments.
Mythos Accessibility is Limited Due to Risks Posed by Its Capabilities
Project Glasswing was created by Anthropic in April 2026 to provide some organizations with an opportunity to access advanced cybersecurity models earlier. At first, the project consisted of major companies representing different industries, including technology, finance, cybersecurity, and infrastructure. Anthropic added about 150 organizations from more than 15 countries to the project.
All organizations willing to access Mythos 5 should meet the security standards. Anthropic restricts access to this tool because it is capable of revealing vulnerabilities and developing exploitable components. According to the company, Mythos 5 is the most powerful tool created by Anthropic for cybersecurity and biological research. Anthropic’s security dashboard detected 1,596 vulnerabilities within 281 open-source projects by May 22. 1,900 candidates were checked by external security firms with a 90.8% true-positive rate. Researchers still have to confirm the results.
Open-Source Findings Create Security Obligations
Payward intends to alert relevant maintainers when Mythos 5 detects valid vulnerabilities in third-party open-source software. The firm has not shared any information on the timeline of such alerts or on any public release of the identified vulnerabilities. In addition, it is unclear how teams will manage projects that do not respond to any security alerts.
Payward has been involved in a previous security issue with regard to its Kraken deposit vulnerability, which was identified using outside research. This incident shows the significance of testing, validating, and communicating during a security investigation process. Anthropic’s requirement for Mythos 5 clients is the thirty-day retention of data for safety monitoring purposes. Payward has yet to clarify what data is subject to this process.
Highlighted Crypto News:
BitMart Faces Aug. 19 Deadline as Users Demand Reserve Disclosure and Repayment Plan
