The alert wasn’t blared out with sirens; however it should probably have been. It seems that in some of the more closed-off policy rooms and on the hurried internal bulletins of European banking and financial oversight bodies, authorities are beginning to suspect something frightening. The upcoming financial apocalypse may not be created by any of our fellow homo sapiens.
First of all, it’s an AI model that can test a system, figure out its vulnerabilities, and in certain conditions, exploit them. Per multiple anonymous industry sources, the ECB has begun reaching out to banks to ask them how they’re feeling about the exposure of this type of new risk. Which is already occurring, and detailed in this article you can learn more about: ECB officials are looking at the risk of the possibility for financial institutions to be exploited by agentic AI models.
Of course, we always talk about the threat of cyber risk from a cyberattack, and of course we’ve always had that. But this isn’t the usual guy in a basement hoodie. This is code with steps it can think through, actions it can chain and in certain tests complex attacks. This is what’s worrying them.
And now, we come to the more bizarre part of this story. Some of those executives have publicly stated that they are “hyper-aware” of the risks; i.e. they don’t sleep too well at night. AI systems similar to Anthropic’s Mythos are reported to have already been able to run autonomous multi-step cyberattack simulation exercises without human intervention. If this doesn’t sound very scary.
But that isn’t the whole picture. There are also reports that Mythos is an early version of AI agents. In general, the evolution of AI is moving from a chat-bot that can generate text, to one that can perform planning and reasoning steps, to AI agents that can actually carry out the plans they generate. Regulators are increasingly worried about AI agents and are discussing what sort of regulations might be necessary.
So where does that leave us? In a sort of half-“amazed,” half-scared place. On the one hand, imagine an AI that can find security holes and fix them before bad actors do. On the other… imagine bad actors finding them first.
Then there’s the matter of trust, which never gets a lot of press. If banks, some of the most risk-averse organizations in the world, are skittish, what should the rest of us think? Do you trust an AI with your money if it has the ability to steal from someone else?
Some people believe an “AI race” is already going on, where nations and companies are scrambling to develop defense as fast as they see offense being used.
In the end, this isn’t just another technology story; it’s one of the many signs of the future.
