The European Central Bank is studying how euro zone banks can protect themselves from cyber threats tied to Anthropic’s Mythos AI model, as officials worry Europe may be exposed to a system it cannot yet examine directly.
Speaking at an event in Spain, Lagarde said the ECB is examining how banks should protect themselves if AI models designed to spot weaknesses in code, including Anthropic’s Mythos, are used by hostile actors, as cybersecurity specialists warn that the same technology meant to improve defenses could also speed up attacks on financial networks.
Mythos access gap leaves Europe exposed
Lagarde said the debate around Mythos has become one of the most closely watched issues in financial cybersecurity, adding that the model’s limited rollout has left Europe at a disadvantage, Reuters reported.
Anthropic has so far restricted access to a small group of U.S. companies, saying the system is being tested carefully because of the risk that its capabilities could be abused.
For European regulators, that creates a difficult challenge: banks may eventually face threats shaped by the model, while supervisors outside the United States remain unable to study it directly.
Lagarde said the ECB is working to map out potential safeguards in case the technology becomes available to state-backed or criminal actors.
“There’s a lot being said about Mythos and it’s a very flashy and sexy topic at the moment,” Lagarde said. “It’s unfortunately one that divides the level playing field between the U.S. and the rest of the world, given that only U.S. companies have been given access.”
ECB pushes banks to prepare for AI-powered attacks
The ECB, which supervises major lenders across the euro zone, has also been asking banks how prepared they are for a new wave of AI-powered cyber tools.
Lagarde said the greatest concern may come from state-sponsored attackers, given the computing resources likely needed to run advanced systems at scale. Her comments point to a broader worry among regulators that AI models built for cybersecurity research could shorten the time needed to find and exploit flaws in banking technology.
“I can assure you that from our perspective as ECB, we are trying to elaborate and identify what defence we should have in place in case this is accessed by a malign state actor or malign actors,” Lagarde said.
The European Commission has also been briefed by Anthropic on Mythos and is reviewing what the model could mean for EU rules and policy.
For regulators, the issue now extends beyond one AI product, with Mythos serving as a test case for how financial authorities prepare for powerful cyber tools that may improve security in trusted hands but create new risks if they spread beyond controlled environments.