Payward, the parent company of cryptocurrency exchange Kraken, has joined Project Glasswing, an initiative run by artificial intelligence company Anthropic. The program is designed to help participating organizations find software vulnerabilities using advanced AI models. Payward will use Anthropic's Claude Mythos 5 model as part of this effort, according to reports from CryptoBriefing and CoinDesk.
Project Glasswing represents Anthropic's push into applied AI security research. The initiative brings together companies from different industries to test how large language models can support vulnerability discovery. Claude Mythos 5 is positioned as a tool capable of analyzing code and systems at scale, potentially surfacing weaknesses that traditional review processes might miss.
For Payward, the decision to join reflects the heightened security demands facing crypto exchanges. Digital asset platforms remain frequent targets for hackers, given the direct financial value stored in custody systems and trading infrastructure. Exchanges have historically relied on manual audits, bug bounty programs, and third-party penetration testing to identify weaknesses before attackers do.
AI-assisted vulnerability hunting adds another layer to that defense stack. Large language models can process vast amounts of code far faster than human reviewers. Proponents argue this speed could help security teams catch issues earlier in the development cycle, reducing the window during which a flaw remains exploitable.
The partnership also illustrates a broader trend of crypto companies working directly with AI developers rather than treating AI tools as generic, off-the-shelf products. By joining a structured research initiative like Project Glasswing, Payward gains access to a model tailored for security tasks, while Anthropic gains a real-world testing ground involving high-value financial infrastructure.
Neither CryptoBriefing nor CoinDesk detailed the specific scope of systems Payward plans to test through Project Glasswing. It also remains unclear how findings from the collaboration will be disclosed publicly, if at all. Companies participating in vulnerability research programs often keep specific technical findings confidential until fixes are deployed, a standard practice across the cybersecurity industry.
Kraken, Payward's flagship exchange, has positioned itself over the years as a security-focused platform within the crypto industry. Adding an AI-driven vulnerability detection layer aligns with that positioning, even as the broader effectiveness of AI tools for security auditing is still being studied across the tech sector.
Market Impact
The announcement is unlikely to move crypto asset prices directly, since it concerns internal security infrastructure rather than trading or custody products. It may, however, influence how other exchanges and crypto infrastructure firms approach AI adoption for cybersecurity purposes.
If AI-assisted vulnerability detection proves effective at Payward, other exchanges could pursue similar partnerships with AI developers. This could accelerate a broader shift toward AI-integrated security practices across the digital asset industry, particularly among platforms handling large amounts of customer funds.
Payward's involvement in Project Glasswing highlights the increasing overlap between AI development and crypto industry security practices. Further details on outcomes from the collaboration may emerge as the initiative progresses.
Frequently Asked Questions
What is Payward?
Payward is the parent company of Kraken, one of the longest-running cryptocurrency exchanges.
What is Project Glasswing?
Project Glasswing is an initiative run by Anthropic that partners with organizations to use AI models for identifying software vulnerabilities.
What is Claude Mythos 5?
Claude Mythos 5 is the Anthropic AI model that Payward will use as part of its participation in Project Glasswing to help detect security flaws.
Does this partnership affect Kraken's trading operations?
The reported partnership concerns security infrastructure and vulnerability detection, not trading products or customer-facing services.