Payward, the parent company of crypto exchange Kraken, has joined Anthropic’s Project Glasswing and gained access to Claude Mythos 5, the AI company’s most capable model for identifying and fixing software vulnerabilities.

Payward said on August 17 that it plans to integrate Mythos 5 into its defensive cybersecurity operations in the coming weeks. The model will be used to scan the company’s operating environments for potential vulnerabilities, with the results feeding directly into its existing security processes.

The initiative follows a US government decision to allow Mythos 5 to be provided to US organizations responsible for operating and defending critical infrastructure. Project Glasswing, launched by Anthropic in April 2026, has since expanded to include organizations across the technology and financial sectors.

For Payward, the project is particularly relevant because digital asset platforms operate continuously and manage sensitive financial infrastructure, including order books, custody systems, and settlement mechanisms. Those systems depend on large and interconnected software stacks, which is why vulnerability scanning can be especially important for teams trying to maintain resilience before issues are exposed in production.

Vulnerability Findings to Benefit the Wider Crypto Ecosystem

Payward said Mythos 5 will be deployed across its environments as part of a proactive effort to identify security weaknesses before attackers discover them. The company views the technology as a way to increase the scale and speed of defensive security work, allowing AI to analyze large volumes of code and infrastructure for potential flaws.

The company also plans to extend the benefits beyond its own systems. If the scans uncover vulnerabilities in third-party or open-source projects, Payward intends to notify the relevant maintainers. The approach is designed to strengthen software dependencies used across the wider cryptocurrency industry rather than limiting security improvements to Payward’s own infrastructure.

The move highlights a growing role for advanced AI models in cybersecurity, particularly as financial and crypto platforms face persistent threats against complex software environments. For companies that rely on widely used open-source components, a finding in one environment can be relevant to many others, which is part of the reason coordinated disclosure and maintainer outreach matter in this kind of program. Anthropic has said Project Glasswing is continuing to expand while the company develops safeguards for broader access to Mythos-class capabilities.

By participating in the initiative, Payward becomes part of Anthropic’s effort to apply frontier AI to the protection of critical software. For the crypto sector, the deployment could also demonstrate how AI-assisted vulnerability discovery can be incorporated into ongoing security programs and contribute to improvements across shared open-source infrastructure.