Security Current Affairs

Kraken's Parent Joins Anthropic's Glasswing to Hunt Vulnerabilities Before Hackers Do

2026-08-18

Kraken's parent company Payward has joined Anthropic's Project Glasswing, giving it early access to Claude Mythos 5, a frontier AI model built specifically to find and patch software vulnerabilities before attackers can exploit them. The announcement came on August 17, and it's a significant move in the crypto sector's race to stay ahead of increasingly sophisticated cyber threats.

Project Glasswing, which launched on April 7, 2026, is Anthropic's structured effort to hand over advanced AI tools to cybersecurity professionals at critical organisations. The catch: these aren't standard Claude deployments. The safety guardrails are deliberately loosened so the AI can reason about attack vectors, probe code for weaknesses, and think like a malicious actor would. That's the whole point.

Payward joins a heavyweight roster already plugged in: Amazon Web Services, Apple, Google, Microsoft, and JPMorganChase are all using Mythos 5. So are roughly 50 other organisations across tech, finance, and now—crucially—critical infrastructure sectors like power, water, and healthcare. Collectively, they've already flagged over 10,000 high or critical-severity vulnerabilities.

The Wyoming-based exchange plans to deploy Mythos 5 across its environments in the coming weeks. Here's where it gets interesting for the crypto ecosystem: any vulnerabilities found in third-party open-source software will be disclosed to project maintainers. That means the broader crypto community benefits, not just Payward itself.

This matters because AI is becoming a two-edged sword. Yes, Payward gets a defensive edge. But the same frontier AI that finds vulnerabilities faster? Attackers are using or will use similar tools to do exactly the same thing. They can automate attacks, craft more convincing phishing campaigns, and scale social engineering. The arms race is accelerating.

Payward's move is ultimately a rebalancing act. By adopting the same frontier models that malicious actors could potentially weaponise, the company is positioning itself ahead of the curve. Co-CEO Arjun Sethi framed it as essential to restoring equilibrium in the security dynamic. You deploy the best tools defensively, or you get left behind.

The U.S. government's decision to make Mythos 5 available to organisations operating critical infrastructure added weight to this play. Payward qualified, and it moved. That's the logic of asymmetric security in 2026.

Source & further reading:

Sources