Payward Partners with Anthropic for Cybersecurity
Payward, the parent company of Kraken, has recently partnered with Anthropic for its cybersecurity initiative known as Project Glasswing. The firm will leverage the power of Claude Mythos 5 to meticulously scan its digital infrastructure for potential security vulnerabilities. This development comes on the heels of an open letter, endorsed by over 40 companies in the Bitcoin and crypto sector, including Kraken, which called on prominent AI research organizations like Anthropic and OpenAI to establish trusted-access initiatives for cybersecurity defenders.
Utilization of Advanced AI Technology
In a statement released on Monday, Payward confirmed it will utilize Mythos 5 for conducting thorough assessments of its systems. Any vulnerabilities identified will be reported back to the company’s security teams for further evaluation. Notably, Payward is the first cryptocurrency entity to participate in Project Glasswing and obtain access to this cutting-edge AI model.
The firm emphasized that their participation enables its security team to gain preliminary access to this advanced technology, greatly enhancing their proficiency in detecting complex software vulnerabilities, thereby safeguarding their extensive customer base globally. Payward also mentioned that any flaws discovered in third-party open-source software will be communicated to the respective maintainers, underscoring their commitment to bolstering the overall security of the crypto ecosystem. This proactive approach could potentially fortify the financial frameworks and enhance the security of open-source code utilized across the industry.
Project Glasswing Overview
Project Glasswing, which was introduced by Anthropic in April and received an expansion in June, offers select vetted organizations access to its most sophisticated cybersecurity models. To date, participants in this initiative have successfully identified thousands of high-priority vulnerabilities.
As the field of artificial intelligence continues to evolve, the capacity of AI models to identify and exploit software flaws is increasingly under scrutiny. A report by Mozilla highlights that Claude Mythos pinpointed 271 vulnerabilities in Firefox just this past April.
Call for Advanced AI Access
The recent letter from the Bitcoin Policy Institute—signed by notable firms such as Ark Invest, Coinbase, Block, and BitGo—urges major AI laboratories to allow qualified security professionals access to their most robust models. This appeal is driven by the recognition that ensuring the safety of open-source financial infrastructure requires advanced AI assistance to preemptively detect and address vulnerabilities before malicious actors can take advantage of them.
Arjun Sethi, Co-CEO of Payward, articulated the inherent imbalance in security dynamics: “In the world of cybersecurity, an attacker only needs to discover one flaw, whereas a defender is tasked with uncovering all potential vulnerabilities continually. Frontier AI changes this dynamic, allowing us to analyze code at machine scale akin to an attacker’s perspective, empowering us to find weaknesses prior to their exploitation.”
