The corporate entity behind the cryptocurrency exchange Kraken, known as Payward, has officially become a participant in the artificial intelligence company Anthropic's "Project Glasswing" program. This move is designed to leverage advanced AI models to strengthen system security, with intentions to share discovered vulnerability information with the open-source community.
Payward announced on August 17 that it will deploy Anthropic's Claude Mythos 5 cybersecurity model within its operating environment over the coming weeks to identify and rectify software flaws. This model represents Anthropic's most sophisticated defensive AI for cybersecurity, capable of analyzing code on a large scale, detecting security defects, and aiding developers in implementing fixes.
The Project Glasswing initiative is a specialized program established by Anthropic, which grants select organizations access to restricted frontier AI models with the goal of safeguarding critical software and infrastructure from cyber threats. The program's membership already encompasses major technology corporations such as Amazon Web Services, Apple, Google, and Microsoft, as well as financial institutions including JPMorgan.
Payward indicated that this collaboration follows the U.S. government's decision to open access to Mythos 5 for American agencies responsible for operating and defending critical infrastructure. The company plans to incorporate the AI model into its existing security framework and will relay vulnerabilities identified in third-party open-source software back to the respective project maintainers.
Arjun Sethi, co-CEO of Payward, noted in a statement that cybersecurity has long suffered from an imbalance between offensive and defensive efforts. Attackers only need to discover a single vulnerability, whereas defenders must race ahead daily to uncover all potential weaknesses. He added that frontier AI technology holds the potential to reverse this asymmetry for the first time.
Currently, artificial intelligence is emerging as a new threat vector for the cryptocurrency industry, as attackers can employ AI to accelerate vulnerability discovery, execute automated attacks, and enhance the sophistication of phishing and social engineering schemes. This development intensifies the security pressures on an industry already frequently targeted by hackers, accelerating the adoption of AI for defensive purposes across the sector.
Payward believes that cryptocurrency trading platforms face security challenges akin to those of traditional critical financial infrastructure. Exchanges, custody systems, and settlement networks operate around the clock, making them highly susceptible to attacks. Anthropic has stated that, as associated safety measures are refined, it plans to progressively broaden access to its Mythos series of cybersecurity capabilities, facilitating wider application in defensive scenarios.
Comments