Payward, the parent company of cryptocurrency exchange Kraken, has joined Anthropic’s Project Glasswing and secured access to Claude Mythos 5, a specialized artificial intelligence model built to identify and remediate software vulnerabilities. The Wyoming-based firm announced the partnership on Monday, stating it will begin deploying the model across its environments in the coming weeks.

Findings from the scans will feed into Payward’s existing security operations, including red-team and blue-team processes, remediation workflows, and its established bug bounty program. The company holds ISO 27001 and SOC 2 certifications and operates trading, custody, and settlement systems that run continuously.

Access to Restricted Frontier Model

Project Glasswing is Anthropic’s initiative that grants selected organizations early use of restricted frontier AI models for defensive cybersecurity. Launched in April 2026, the program has included major technology firms such as Amazon Web Services, Apple, Google, Microsoft, Cisco, Nvidia, and Palo Alto Networks, along with financial institutions including JPMorganChase. Access has since expanded to additional technology and finance participants.

Claude Mythos 5 is Anthropic’s most capable model for defensive cybersecurity work. It is designed to analyze code at scale, detect vulnerabilities, and assist in developing fixes. Anthropic has limited distribution because the same capabilities that help defenders can also support the creation of exploits. Availability has been aligned with U.S. government decisions permitting use by American organizations that operate or defend critical infrastructure.

Payward stated that its participation follows the U.S. decision to allow Mythos 5 access for entities securing critical infrastructure. The company maintains that cryptocurrency platforms face security demands comparable to other critical financial systems, given their continuous operation and appeal as targets for attackers.

Deployment Plans and Broader Impact

Payward plans to scan all of its environments. Human review will remain part of the process to validate findings before any classification or remediation. The company has not published specific targets, a detailed deployment timeline, or performance metrics for the initial scans.

Vulnerabilities identified in third-party open-source software will be reported to the relevant project maintainers through responsible disclosure. Payward noted that this step could strengthen shared infrastructure used across the cryptocurrency sector, as many platforms rely on common open-source packages.

Artificial intelligence has increased both offensive and defensive pressures in the industry. Attackers can use AI tools to locate weaknesses more quickly, automate campaigns, and improve social-engineering efforts. Defenders are responding by adopting similar technology to reverse the traditional imbalance.

Anthropic has indicated it intends to expand access to Mythos-class capabilities as additional safeguards are developed for wider distribution. Partners in Project Glasswing have previously surfaced thousands of high- and critical-severity issues across widely used software, with independent triage remaining a key step in the process.

Payward is the first reported cryptocurrency firm to gain formal entry into the program. The company operates additional businesses including NinjaTrader, Breakout, xStocks, Bitnomial, and CF Benchmarks. Its second-quarter adjusted revenue reached $508 million, a 17 percent increase from the prior year.

The announcement arrives as AI models continue to demonstrate strong performance on cybersecurity benchmarks while also raising dual-use concerns that keep the most advanced systems restricted to vetted participants.