- Published on
Kraken Parent Payward Joins Glasswing, Gets Access to Claude Mythos to Hunt Security Flaws
- Authors
- Name
- DeepHorus
Kraken parent Payward has become the first crypto company to join Anthropic’s Project Glasswing, gaining access to Claude Mythos 5 for cybersecurity defenses. The company announced Monday that it will use the frontier AI model to scan its systems for vulnerabilities, with findings routed to its security teams for review and patching. Payward said the partnership sharpens its ability to combat sophisticated software flaws and protect millions of customers worldwide.
The move follows an open letter sent last week by over 40 Bitcoin and crypto companies, including Kraken, Coinbase, Block, and BitGo, urging leading AI labs to open trusted-access programs to qualified defenders. Organized by the Bitcoin Policy Institute, the letter argued that developers protecting open-source financial infrastructure need frontier AI to find and fix vulnerabilities before attackers exploit them. Project Glasswing, launched by Anthropic in April and expanded in June, gives vetted organizations access to its most capable cyber models.
According to Payward, Mythos 5 will scan its systems and also examine third-party open-source software. Any flaws discovered in external projects will be reported to maintainers, potentially strengthening the broader crypto ecosystem. The company emphasized that this approach could protect its financial infrastructure while improving open-source software used across the industry.
The news comes amid growing evidence that AI models are powerful tools for both finding and exploiting vulnerabilities. In April, Mozilla reported that Claude Mythos identified 271 vulnerabilities in Firefox during testing. Anthropic’s Glasswing partners have already uncovered thousands of high- or critical-severity vulnerabilities since the program expanded.
Payward Co-CEO Arjun Sethi framed the development as reversing the defender’s disadvantage. “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,” he said. He added that frontier AI can read code at machine scale, allowing defenders to find flaws before attackers build exploits.
Anthropic did not immediately respond to a request for comment. The partnership signals a growing trend of crypto firms seeking frontier AI access for defensive security, likely pushing more AI labs to formalize trusted-access programs. As threats evolve, the integration of advanced AI into financial security may become standard practice for major exchanges.
