When AI Insiders Blow the Whistle: How Worried Should We Be?

In just a few years, AI systems have gone from narrow tools to general-purpose models that can write code, conduct research, and reason across complex domains. Leading forecasts now place AI matching or exceeding humans at most cognitive work somewhere between 2030 and 2040. Timelines remain uncertain, but the direction is clear: capabilities that seemed decades away are arriving much sooner than expected.

AI of this kind would not be just another general-purpose technology. It could reshape labour markets, scientific progress, military capability, and the balance of power between states. Alongside the opportunities come risks our institutions were not designed for: large-scale misuse, accidents in critical systems, concentration of power, systems whose behaviour we cannot reliably predict, and the loss of human control over them, which many researchers see as an existential threat.

Europe lacks the strategic awareness, institutional capacity, and integrated strategy this transition demands. There is no Draghi-scale plan for AGI preparedness. But Europe still has serious assets: a strong research base, regulatory credibility, and a record of building institutions when it has to. The window is narrow, but it is still open.