The Kraken parent company Payward disclosed that it had joined Project Glasswing, initiated by Anthropic, and obtained access to its security model Claude Mythos 5. The model will be used to identify and repair software gaps and access internal company security processes in the coming weeks.
Project Glasswing covers multiple institutions
Project Glasswing is an Anthropic project for key software and infrastructure security that allows some agencies to use restricted front-line AI models to address cyberattack risks. Public information indicates that participants have included institutions such as Amazon Cloud Technologies, Apple, Google, Microsoft and Morgan Chase.
Payward stated that if a loophole was found in the third-party open source software, the company would disclose the issue to the counterpart project maintainer. This means that, in addition to being used for internal repairs, associated discoveries may also help to reduce risk in the wider ecology of encryption software.
Commence internal deployment in the coming weeks
According to the company, Claude Mythos 5 will be used for large-scale code analysis, gap identification and repair support and will incorporate the results into existing security plans. According to Payward, encryption platforms are similar at the security level to other key financial infrastructure, and trading, hosting and clearing systems require all-weather operations and are therefore a long-term focus of attacks.
According to Payward, the United States Government had previously opened Mythos 5 to United States agencies responsible for the operation and protection of critical infrastructure. Anthropic launched Project Glasswing in April 2026, then expanded the project to include more organizations in the fields of science, technology and finance.
AID's synchronous acceleration.
The article mentions that AI is increasing its ability to both attack and defend. The attackers can detect loopholes more quickly, automate part of the attack and make fishing and social work attacks more confusing. This has also prompted encryption companies to use similar tools more quickly for defence purposes.
The CEO Arjun Sethi, Co-Chair of Payward, stated that there was a long-standing marked asymmetry in security protection, that only one gap needed to be identified by the attackers and that the defensive parties needed to continue to detect the problem in advance. He thinks the front-line AI is changing the situation.
Additional information:Anthropic states that, as security measures improve, future plans are to open Mythos-level network security capabilities to more agencies.