How to Verify Kraken Developments Across Three Sources

Three independent publishers reported two confirmed Kraken developments in the same week. According to Cointelegraph, Kraken parent Payward booked $508 million in Q2 adjusted revenue, up 17% year over year, while spot trading volume fell. According to CoinDesk and Decrypt, Payward also joined Anthropic's Project Glasswing, gaining access to a cybersecurity AI model that Decrypt identifies as Claude Mythos.

Together these Kraken developments change how the exchange runs day to day: its security workflow gains an AI-assisted scanning step, and a growing share of its revenue now comes from outside transaction-based activity.

Kraken Developments Before you read the reports what is already confirmed

Treat each publisher development as separate evidence. The Glasswing story appears twice, from CoinDesk and Decrypt, and both name the same actor and action: Payward joins a program operated by Anthropic that gives vetted organizations access to its cybersecurity AI. CoinDesk adds that Payward will use the model to hunt vulnerabilities and share open-source findings.

Decrypt's version names the model as Claude Mythos and describes the program as limited to vetted organizations, which makes this a permissioned change rather than an open product release. The financial story stands on its own from Cointelegraph: revenue up 17 percent, funded accounts up 42 percent, spot volume down, and more revenue arriving from non-transactional sources.

These three data points are the ceiling of what is confirmed. Anything beyond them, including deployment timelines and revenue-line detail, is inference until a source document says otherwise.

Step 1: Separate the Glasswing membership from the revenue shift for Kraken Developments

The two storylines answer different questions. The Glasswing membership is an operational security change: Payward's vulnerability discovery moves from purely internal tooling to AI-assisted scanning, with results partly published as open-source findings that outside projects can use. The affected participants are Payward as the joining organization, Anthropic as the program operator, and Payward's security engineers as the internal users of the tool.

The Q2 result is a business-mix change. According to Cointelegraph, Payward grew revenue despite weaker crypto spot activity because funded accounts jumped 42 percent and a growing share of revenue came from outside transaction-based activity. The affected participants here are Payward's account holders and the non-transactional product lines absorbing that revenue share.

Step 2: Name the operating impact for each participant for Kraken Developments

For Payward's security engineers, the operating impact is procedural: vulnerability hunting gains a new detection capability and a publication commitment. The open-source commitment means some findings leave the company entirely, which benefits external maintainers who receive them.

For the financial side, the impact is structural rather than directional. A company earning 17 percent more while its core spot venue handles less volume means service load is rising from funded accounts, while fee income depends less on trading activity. That combination affects capacity planning and how revenue reporting is read: non-transactional lines now carry enough weight to move the total.

Both threads describe the same thing from different angles: an exchange operator adjusting how it runs security and how it earns, without either change being a market call on any token price.

Step 3: Check the risks that weaken this reading for Kraken Developments

Several limits apply before treating these reports as complete. The 17 percent revenue figure and 42 percent funded-account growth are Q2 measurements reported by Cointelegraph; they describe one quarter, not a durable trend, and quarterly results are subject to volatility in crypto spot activity that can reverse quickly.

The Glasswing report confirms membership but not scope. Access to a cybersecurity AI model does not remove operational-error risk, smart-contract or network risk, or custody risk from the exchange's broader business. Security tooling reduces some attack surface at best; it does not eliminate it, and no source claims otherwise. Readers should also note that secondary reporting, even from three outlets, is not the same as a filing or program document.

Step 4: Run the source-document check for Kraken Developments

Three items remain unverified in the supplied reports. First, the scope and terms of Claude Mythos access, which belongs in Anthropic's official Project Glasswing program documentation. Second, the timeline and criteria for publishing open-source security findings, which belong in a Payward security disclosure policy if one exists. Third, the exact composition of non-transaction revenue, which belongs in Payward's own Q2 filing.

The concrete next step is to open those documents and match each claimed detail against them. If a detail cannot be traced to the program terms or the filing, treat it as unconfirmed and stop there rather than extending the reading.

When not to proceed with a conclusion for Kraken Developments

Hold the analysis at the confirmed level if any check fails: no Glasswing program page exists, no Payward filing discloses the revenue breakdown, or the published figures differ from the reported ones. Until those documents appear, the Glasswing membership and the Q2 headline numbers are the floor of what is known, and the scope of Claude Mythos access, the findings publication schedule, and the non-transactional revenue mix remain open questions. That boundary, not speculation, is where this reading should rest.

Reference

• https://www.coindesk.com/tech/2026/08/17/kraken-parent-payward-joins-anthropic-s-project-glasswing-for-ai-security-push
• https://cointelegraph.com/news/kraken-parent-payward-revenue-rises-as-trading-volume-falls-q2
• https://decrypt.co/375836/kraken-payward-anthropic-project-glasswing-claude-mythos