Kraken Parent Payward Deploys Anthropic’s Claude Mythos 5 to Audit Crypto Infrastructure
Crypto firm becomes first in the industry to gain defensive access to Anthropic's frontier AI models
Payward, the parent firm of digital asset exchange Kraken, has become the first cryptocurrency business to secure defensive access to Anthropic’s Claude Mythos 5 model under the AI lab’s Project Glasswing cybersecurity initiative.
The participation follows a joint letter issued last week by more than 40 cryptocurrency and Bitcoin companies—including Kraken—to Anthropic, OpenAI, and other top artificial intelligence research facilities, urging them to grant certified security professionals access to defensive AI tooling.
Payward announced Monday that it will deploy Mythos 5 to analyze its corporate software infrastructure for technical weaknesses, routing all detected flaws to internal security analysts for manual review. The company marks the initial digital asset platform admitted to Project Glasswing with active access to the Mythos 5 model.
“Selection gives Payward’s security division early access to the same class of model, sharpening its ability to combat sophisticated software vulnerabilities and protect millions of customers across the globe,” Payward stated.
Security vulnerabilities discovered within external open-source dependencies will be disclosed to project maintainers, Payward explained, noting that the workflow is designed to fortify its primary financial systems while bolstering shared code libraries relied upon across the digital asset market.
Digital asset networks and crypto exchanges frequently incorporate open-source financial infrastructure, where minor software bugs or smart contract errors have historically resulted in severe security breaches and substantial capital losses. By deploying advanced language models capable of reasoning through complex code dependencies, defenders aim to match the speed of automated exploit scanners used by threat actors.
Anthropic established Project Glasswing to supply approved defensive organizations with early access to its specialized cybersecurity models. Originally introduced in April and broadened in June, the initiative’s participating partners have identified thousands of critical- and high-severity software flaws to date.
Anthropic did not immediately respond to a request for comment.
The deployment occurs alongside mounting data indicating that frontier AI architectures possess capabilities for both discovering and exploiting software flaws. In April benchmarks, Mozilla documented that Anthropic’s Claude Mythos uncovered 271 security flaws within the Firefox web browser.
The recent industry petition—coordinated by the Bitcoin Policy Institute and backed by firms including Coinbase, Ark Invest, Block, and BitGo—emphasized that security teams safeguarding open-source monetary systems must utilize cutting-edge AI models to identify software defects prior to malicious exploitation.
“Security has always been an unfair game. An attacker needs to find one flaw. A defender has to find all of them, first, every single day,” Payward Co-CEO Arjun Sethi said. “Frontier AI is the first thing that flips that asymmetry. A model can read every line of code the way an attacker would, at machine scale, so we find the flaw before anyone can build the exploit.”
