The Bundesbank Warns of the Risks of AI Mythos
GAI
AI
UTED
MYTH
Generative AI is now at the heart of debates on global financial stability. Joachim Nagel, president of the Bundesbank, has issued a severe warning: Anthropic’s Mythos model could expose European banking systems to unprecedented risks. Will Europe be able to respond before it’s too late?
In Brief
- Bundesbank president Joachim Nagel has warned Europe about Anthropic’s Mythos AI model.
- Mythos can detect and exploit vulnerabilities in banking software, representing a major risk to financial stability.
- Nagel calls for equal access to this technology for European institutions.
- Europe has only 3 advanced AI models in 2024, compared to 40 for the United States.
A Double-Edged Technology That Worries Frankfurt
This Tuesday, in Rome, Joachim Nagel spoke before his European counterparts. The president of the Deutsche Bundesbank, and a member of the ECB Governing Council, clearly identified Mythos, the new AI model developed by Anthropic, as a potential threat to the stability of the European financial system.
His observation is straightforward: Mythos is capable of quickly identifying and exploiting vulnerabilities in the software used by banks. Useful to strengthen digital defenses, certainly. But formidable in the wrong hands.
“The use of AI in the financial sector paves the way for new and sophisticated cyber risks,” he warned. “Autonomous AI agents might adopt malicious behaviors.”
His question, posed without hesitation, sums it all up: can we ask European banks to protect systems they are not even authorized to fully audit?
A European AI Delay That Worsens Risks
The numbers speak for themselves. In 2024, the United States developed 40 advanced AI models, China 15, and Europe… only 3. On the private investment side, the gap is just as stark:
- United States: $109.1 billion
- Europe: $19.4 billion
- China: $9.3 billion (offset by $62 billion in public investments)
In this context, Nagel demands that European banks and regulators have access to Mythos before the model reveals further vulnerabilities in their infrastructures.
“All relevant institutions should have access to this technology to avoid any competitive distortion,” he emphasized.
Meanwhile, Anthropic continues to strengthen its dominant position. The company has just signed a colossal deal with Amazon: more than $100 billion invested in AWS over ten years, with guaranteed computing capacity up to 5 gigawatts to train its models.
Europe faces an uncomfortable reality: it depends on technologies it does not control to protect systems it cannot fully audit. Without quick access to tools like Mythos, the continent risks not only widening its technological gap but also leaving its banks exposed to tomorrow’s cyber threats.
