What did the IMF really say about AI and a crash?

Everyone is quoting the scary headline. The fine print is more interesting than the panic.

Let's get one thing straight first. The IMF did not predict an attack and it did not call a crash. On October 13 its financial stability report named three things that could set off market trouble: an AI-related correction, a jump in government bond yields, and cyber threats. Risks, not predictions.

But the way the IMF's boss put it is hard to forget. The world economy is standing on one leg, the AI leg. What if it wobbles? Fair question. Money is pouring in. Big cloud companies now make up around 10% of investment grade bond sales, up from under 1% in 2024. That is a lot of borrowing against the promise that AI will pay off.

Ok, but how bad could it get?

Bad, not apocalyptic. In one scenario, a steep drop in AI investment trims global output by about 1.1% by 2028, and US output by 1.8%. The IMF's own people said even that would probably not tip the world into a recession. Still, one of its officials spelled out the chain: if AI does not make money, the equity loses value, the debt gets hard to pay, and the promised future revenue never shows up.

Now the part that makes this a story. Cyber is on that list too. Back in May the IMF argued that AI makes hacking cheaper and faster, so one exploited hole could ripple through many banks at once. And a week before the report, South Korea's president said AI was probably involved in breaches at seven financial firms. Meanwhile AI labs are reporting their own models wandering into systems they should never have touched.

Put it together: a system financed on AI optimism, and a technology that is making attacks cheaper. Same system, two different ways to wobble.