Kraken's parent Payward is using Anthropic's Claude Mythos 5 AI to scan for software vulnerabilities, after a U.S. government clearance. The earlier version found 10,000+ critical flaws.
The parent company of cryptocurrency exchange Kraken, Payward, has joined Anthropic's Project Glasswing security initiative and started using the Claude Mythos 5 model to find vulnerabilities in its code base. The announcement came via a tweet from Payward on August 17, with co-CEO Arjun Sethi saying the AI model shifts the advantage toward defenders.
"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," Sethi said in a statement.
Mythos 5 is Anthropic's most capable model for defensive cybersecurity. It can analyze code at large scale, detect weaknesses, and suggest fixes. Anthropic controls access tightly because the same technology could be used to build exploits. The U.S. government cleared Mythos 5 for distribution to entities that protect critical infrastructure. Payward's authorization followed that determination.
Payward's infrastructure powers Kraken's digital asset trading, custody, and settlement operations, which run around the clock. The company said crypto platforms face security threats similar to those of other mission-critical financial systems.
Results from the AI scans will feed into Payward's existing security workflows. The company did not say whether Mythos 5 will examine live production environments or isolated code repositories. All AI-flagged issues will go through human verification before being confirmed as real vulnerabilities. Complex software analysis by AI often produces false positives, the company noted.
The Ethereum Foundation found similar results during its own AI-assisted security reviews. Vulnerability reports from AI still need thorough human review, especially for sophisticated protocol implementations, the foundation's research showed.
Payward also committed to sharing confirmed vulnerabilities found in open-source third-party software with the project maintainers. It has not released details on disclosure procedures or timelines.
Anthropic launched Project Glasswing in April 2026. Early participants included Amazon Web Services, Apple, Cisco, Google, JPMorganChase, and Microsoft. The program has since grown to roughly 150 organizations across more than 15 countries. Participants must meet strict security criteria before gaining access.
An earlier version of Mythos, called Mythos Preview, found more than 10,000 high-severity or critical vulnerabilities in commonly used software packages. Anthropic's public dashboard showed a 90.8% true-positive accuracy rate, though only 97 documented vulnerabilities had received upstream patches by May 2026.
Organizations using Mythos 5 must agree to Anthropic's 30-day data retention policy, designed for safety oversight. Payward has not specified which code repositories or system data will be transmitted during scanning.
The company has not released performance metrics or targets. Scan results will supplement its current security framework rather than trigger automatic code changes, Payward said.
Kraken is one of the largest crypto exchanges by volume and ranks among the top platforms for security. Traders evaluating platforms can review AlphaScala's guide to best crypto brokers.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.