> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Kraken's Parent Company Deploys AI Agent to Hunt Bugs That Could Drain Billions
- URL: https://wire.fourthweb.ai/krakens-parent-company-deploys-ai-agent-to-hunt-bugs-that-could-drain-billions/
- Published: 2026-08-17T20:47:49.000Z
- Updated: 2026-08-17T22:32:41.000Z
- Description: The exchange that holds billions in customer crypto just deputized an AI agent to hunt for the bugs that could drain it all. Payward, Kraken's parent company, joined Anthropic's Project Glasswing, an initiative using Claude Mythos 5 to find software vulnerabilities before attackers do
- Author: Travis Wright
- Tags: Real World Assets, AI Agents, Smart Contracts, Anthropic

**The exchange that holds billions in customer crypto just deputized an AI agent to hunt for the bugs that could drain it all.**

### The Summary

- [Payward, Kraken's parent company, joined Anthropic's Project Glasswing](https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push?ref=wire.fourthweb.ai), an initiative using Claude Mythos 5 to find software vulnerabilities before attackers do
- [The move signals growing reliance on AI for proactive cybersecurity](https://cryptobriefing.com/payward-anthropic-glasswing-claude-mythos-5/?ref=wire.fourthweb.ai) in protecting digital assets
- [Payward will share its vulnerability findings as open-source intelligence](https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push?ref=wire.fourthweb.ai), contributing to collective defense across the industry

### The Signal

[Payward's participation in Project Glasswing](https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push?ref=wire.fourthweb.ai) marks a turning point in how crypto exchanges think about security. Instead of waiting for whitehat researchers or, worse, finding out about vulnerabilities when funds disappear, they're deploying an AI agent specifically trained to think like an attacker. Claude Mythos 5 doesn't just scan for known patterns. It reasons about how systems break.

This matters because crypto security is usually reactive. An exploit hits, millions vanish, then everyone patches the same vulnerability. [Payward is flipping that model](https://cryptobriefing.com/payward-anthropic-glasswing-claude-mythos-5/?ref=wire.fourthweb.ai) by hunting bugs at machine speed before they become headlines.

> "[AI agents](https://wire.fourthweb.ai/tag/ai-agents/) don't get tired, don't miss edge cases in 3am code reviews, and don't need bug bounty incentives to stay motivated."

The open-source commitment is the surprising part. Most exchanges treat security findings like trade secrets. [Payward says it will publish what Claude Mythos 5 discovers](https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push?ref=wire.fourthweb.ai), raising the defensive floor for everyone. That's either altruism or pragmatism: if every exchange patches the same critical bugs, the entire ecosystem becomes a harder target. Attackers follow the path of least resistance. Remove the easy paths, and they move to softer targets outside crypto entirely.

Project Glasswing itself is [Anthropic](https://wire.fourthweb.ai/tag/anthropic/)'s bet that cybersecurity is an AI-native problem. Human security researchers are scarce, expensive, and slow. Claude Mythos 5 can audit codebases 24/7, flag suspicious patterns in [smart contracts](https://wire.fourthweb.ai/tag/smart-contracts/), and simulate attack vectors faster than any red team. The model isn't replacing security engineers. It's doing the grunt work so humans can focus on the judgment calls: which risks matter most, how to communicate fixes without tipping off attackers, and whether a vulnerability is theoretical or actively exploitable.

### The Implication

If you run infrastructure that touches crypto, AI security agents just became table stakes. Payward isn't doing this for headlines. They're doing it because the cost of a breach, both financial and reputational, is higher than the cost of letting an AI agent crawl through their stack. Watch for other exchanges to either join Glasswing or spin up their own AI red teams in the next six months.

For security researchers, this is clarifying. The low-hanging fruit, the bugs that pattern-matching can catch, those are now AI work. The valuable human skill is adversarial creativity: imagining the attack no one has thought of yet, the one that requires combining three unrelated vulnerabilities in a novel way. If an AI can find it by brute reasoning, it's no longer a premium skill.

### Sources

[CoinDesk](https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push?ref=wire.fourthweb.ai) | [Crypto Briefing](https://cryptobriefing.com/payward-anthropic-glasswing-claude-mythos-5/?ref=wire.fourthweb.ai)