Kraken Parent Payward Joins Anthropic’s Project Glasswing for AI Bug-Hunting

Payward, the Wyoming-based parent company of crypto exchange Kraken, has joined Anthropic’s Project Glasswing, gaining access to Claude Mythos 5, an AI model built specifically to find and patch software vulnerabilities before attackers can exploit them. The company confirmed the move on Monday, positioning itself alongside a short list of firms Anthropic has vetted to run its restricted frontier security model.

Project Glasswing, launched by Anthropic in April 2026, already counts Amazon Web Services, Apple, Google, Microsoft and JPMorganChase among its participants. Payward’s inclusion marks one of the first times a crypto-native company has been given access to the program, which restricts Mythos 5 to organizations Anthropic considers capable of securing critical software and infrastructure.

What Payward plans to do with Mythos 5

Payward said it intends to deploy Mythos 5 across its own environments in the coming weeks, feeding the model’s findings directly into its existing security program. The company also plans to disclose any vulnerabilities it uncovers in third-party open-source software to the relevant project maintainers, a step that could extend the benefit beyond Kraken’s own infrastructure to the wider crypto ecosystem that relies on shared open-source code.

Access to Mythos 5 followed a decision by the U.S. government to make the model available to American organizations that operate and defend critical infrastructure, according to Payward. Anthropic has been steadily widening Project Glasswing’s participant base since its April 2026 launch, moving from a small circle of technology giants into finance and, now, crypto.

Why crypto exchanges are racing to adopt defensive AI

The partnership lands as AI tools increasingly cut both ways in cybersecurity. The same capabilities that let defenders scan code for flaws also give attackers faster ways to discover exploits, automate intrusion attempts, and craft more convincing phishing and social-engineering campaigns. Crypto platforms, which sit on liquid, often irreversible assets, remain among the most frequent targets for these attacks.

That dynamic is fueling a parallel race among exchanges and custodians to put the same class of AI models to work on the defensive side before adversaries can weaponize them at scale. By joining Glasswing, Payward is betting that early access to a frontier model like Mythos 5 gives its security team a head start on vulnerabilities that could otherwise surface first in an attacker’s toolkit.

Why it matters for Kraken users

For Kraken’s customers, the practical upside is a security program that now includes automated vulnerability scanning powered by one of the industry’s most closely guarded AI models, rather than relying solely on traditional audits and bug bounties. The commitment to disclose third-party open-source flaws to maintainers also signals a defensive posture that extends past Payward’s own perimeter, a detail worth watching as more exchanges weigh similar partnerships with AI labs in the months ahead.

Read more: SafePal Breach Exposes 40,000 Addresses, Pushing Weekly Wallet Leak Past 53,000

Sources

• decrypt.co
• coindesk.com