Claude Mythos introduces Crack of Doubt for banking customers
For decades, banks have been the ultimate trusted space. We assume they’ve got the fraud stuff handled, mostly because we’re told scams are our fault with social manipulation that happens outside their walls. But Anthropic just dropped a massive rock into that calm pond.
They announced Claude Mythos, an AI so powerful at coding and logic that it can spot security flaws in almost every major OS before anyone else even knows they exist. Anthropic found it could identify bugs that had been sitting undiscovered for 27 years. The industry response wasn't a press release, it was a panic.
On April 8, US Treasury Secretary Scott Bessent and Fed Chair Jerome Powell summoned the CEOs of JPMorgan Chase and Goldman Sachs to an emergency meeting. They weren't talking about interest rates; they were talking about a systemic threat to the global financial system.
Holly molly
The market didn't wait for the meeting to end. As the news of the Mythos leak spread, investors wiped billions off the valuations of cybersecurity and enterprise software giants in a matter of days. Salesforce stock alone dropped over 8%, hitting its lowest level in years, while security firms like CrowdStrike saw their shares tumble as much as 11%. The market was pricing in a terrifying new reality. If an AI can find vulnerabilities that humans missed for decades, the old defensive business models are suddenly looking very flimsy.
Europe was just as spooked. The European Central Bank (ECB) started quizzing bankers about their preparedness, calling Mythos a credible AI Hazard. In Germany, the financial watchdog BaFin and the Bundesbank were in emergency huddles with the German Banking Association. While Deutsche Bank CEO Christian Sewing was publicly trying to play it cool, saying there are no "alarm bells," his team was forced to set up a special working group just to help smaller banks survive the fallout.
Anthropic is so worried they won’t even release the model. They’ve locked it in Project Glasswing, giving only a few firms like Goldman Sachs and Microsoft access to patch the holes before the hackers find them. This is Anthropic essentially saying ‘Sorry, we built this. Good luck’.
When the Fed Chair, the ECB, and the Bundesbank are all in a room demanding answers, the rest of us collectively shit-a-brick. For banking customers this is a crack of doubt moment.
The crack of doubt
A crack of doubt is the beginning of the end. It is subconscious moment where you realise the relationship is not what it used to be. These doubts build as questions go unanswered. The product isn't achieving what its foundational obligations were. In this case, keep my money safe. Customers haven't left the bank yet, but the questions are getting louder. Is my money actually there? Is a bot about to empty my life savings? Should I store my money under the bed?
When trust plummets like this, people ramp up the questions and start thinking about the alternative providers.
Instead of staying silent, banks need to start talking. They need to prove the floor isn't about to fall out. If they don't, that crack of doubt is going to lead to a lot of people seeking the next phase of their off-boarding journey - Acknowledgement. That stage is harder to talk your customers away from.
