Deutsche Bank's CEO thinks German banks are ready for AI-powered hacking, which means they probably aren't.
The Summary
- Deutsche Bank CEO Christian Sewing says German banks are "well-prepared" for cyber risks from Anthropic's new AI model, which has sparked global concerns about AI-enabled hacking
- Banks claiming readiness for AI threats before those threats fully materialize is the corporate equivalent of saying "we've got this" right before everything breaks
- The fact that a major bank CEO feels compelled to publicly assure everyone they're not panicking suggests there's something worth panicking about
The Signal
Anthropic just shipped something called Mythos that has global banks nervous enough that Deutsche Bank's chief executive is doing press rounds to say everything's fine. That's the tell. When bank CEOs start proactively claiming they're not worried, start worrying.
The technical reality is straightforward. Large language models are getting better at code, which means they're getting better at finding vulnerabilities in code. Every security system is just code someone wrote. Every firewall, every authentication layer, every encryption protocol. If an AI can read code well enough to understand what it does, it can read code well enough to understand what it doesn't do.
"German banks are well-prepared for heightened cyber risks from AI models that could enable sophisticated hacking."
Here's what's actually happening. Banks have been fighting human hackers who need to sleep, who make mistakes, who get bored. AI agents don't sleep. They don't get bored. They can probe a system a million times in the time it takes a human to probe it once. They can learn from every failed attempt instantly. The game just changed and the referees are still explaining the old rules.
The interesting part isn't whether banks are ready. They're not. No one is. The interesting part is what this means for the agent economy everyone's been building. We've spent two years talking about AI agents that book your flights and summarize your emails. Anthropic just reminded us that agents can also dismantle your security infrastructure while you sleep.
The implications break three ways:
- Defensive AI becomes mandatory infrastructure, not optional tooling
- Every company building agents now needs a red team running adversarial agents against their systems
- The cybersecurity market just became an AI arms race with real stakes
The Implication
If you're building in Web4, this is your warning shot. Every agent you ship is also a template for what a hostile agent could do. The companies that survive the next five years will be the ones that assumed bad actors would use their exact same tools, but faster and meaner.
Watch what the banks actually do, not what they say. If Deutsche Bank quietly triples their AI security budget in the next quarter, that's your signal. When executives say "we're not panicking," check their procurement records.