ECB’s Lagarde warns Europe against leaning on foreign AI models

Europe has spent years worrying about its energy dependence. Christine Lagarde would like it to start worrying about its AI dependence too.

In a speech in Vienna on September 14, 2026, the European Central Bank president warned that Europe’s growing reliance on artificial intelligence models built in the US and China leaves the continent exposed. Her point was simple: if someone else owns the tools, someone else can take them away.

“A withdrawal of access…would reach every sector at once.”

What Lagarde actually said

Lagarde laid out a list of areas where imported AI is already doing real work. These include border screening, tax administration, transportation, healthcare and banking payments.

She compared the risk to past episodes where outside forces destabilized local economies. The prescription that followed was a push for Europe to build its own AI capacity, expand data center and compute infrastructure, and secure a meaningful role in the AI supply chain.

Lagarde returned to the theme on October 1, 2026, at a European Systemic Risk Board conference. This time the focus shifted toward financial stability, with AI flagged as a potential source of market volatility and cybersecurity threats.

As of October 2026, nearly 90% of significant euro-area banks use generative AI. When almost every major bank leans on the same category of technology, a disruption stops being an IT problem and starts looking like a systemic one.

The numbers behind the warning

Europe holds about 5% of global AI computing capacity. The US holds approximately 75%.

In 2025, the US produced 59 notable AI models and China produced 35. Europe produced 2, one from France and one from the UK.

Technically, the UK is not in the EU anymore. So the bloc’s own tally is even thinner than that already modest figure suggests.

Europe’s data center capacity deficit is set to grow more than sixfold over the next decade. Closing that gap could cost up to €600 billion.

Rapid AI adoption could lift productivity by up to 4% over a decade, according to the figures cited alongside Lagarde’s remarks.

Why this went from theoretical to practical

In June 2026, a US export-control action temporarily suspended European users’ access to two advanced AI models. The suspension was temporary, but the precedent was set.

It showed that access to frontier AI can be switched off by policy decisions made outside Europe. Users had no say in the matter and limited fallback options.

It also explains why a central banker, rather than a tech minister, is making this argument. The ECB’s mandate centers on price stability and the health of the financial system. Payments and banks sit squarely inside that remit, and both now run partly on imported AI.

What this means for Europe and its markets

For European policymakers, Lagarde’s remarks add institutional weight to the sovereignty push. When the head of the ECB frames AI dependence as a stability risk, it becomes harder to treat compute investment as an optional industrial policy experiment.

Banks face a more delicate balance. Nearly 90% of significant euro-area lenders already use generative AI, and the productivity gains are real. But supervisors now have an explicit signal that concentration on a handful of foreign models is a risk they intend to watch.