The arrival of Claude Mythos, the new advanced artificial intelligence model developed by Anthropic, scares European banks, which risk serious problems in terms of cybersecurity. For this reason, the ECB has called a meeting for tomorrow with the heads of the euro area credit institutions, to jointly address the systemic risks linked to the spread of Mythos.
What is Mythos and why is it a concern
Mythos is Anthropic’s next-generation AI model, believed by cybersecurity experts to radically transform both defense and offense in financial cyberspace. According to the ECB, the model’s high ability to identify flaws in the IT systems of banks, financial infrastructures and payment networks could facilitate cyber attacks and “break through” even the most sophisticated protection systems. The model is considered by supervisors to be a significant challenge for the banking industry and its technological systems, so much so that it has triggered a series of warnings from regulators and policymakers.
The meeting with the banks
The ECB therefore called the meeting with bankers – writes the Financial Times – to push banks to accelerate a solution, including the need to share information with US banks equipped with more advanced technologies.
Frank Elderson, vice-president of the ECB’s banking supervisory arm, warned that given the speed of cyber threats powered by AI, they need to be “dealt with more quickly” than in the past. Indeed, when “one of the big software providers releases a patch”, it is now possible to act “not in weeks, but perhaps in 30 minutes”. “This means that, once the patch has been published – Elderson explained – a bank must have processes in place to ensure that it applies these patches much faster than is currently the case of the market”.
The technological gap is widening
A peculiar element aggravates the European scenario: unequal access to technology, the diffusion of which has for the moment been inhibited to the general public and made available to US banks, to test the related protection systems, not also to European banks.
Anthropic has in fact allowed access to Mythos to a limited number of institutes, as part of the test called Project Glasswing. Big U.S. banks, which got early access to Mythos, are racing to fix dozens of data system vulnerabilities reported by the tool, while Japan’s three largest institutions are expected to get the green light in about two weeks.
The EU risks being left behind. In this regard, Elderson stressed that the lack of access to Mythos by euro area banks increases the severity of the problem, adding that “lack of access is no excuse for inaction”.
Risk analysis
The concern is not limited to the cyber aspect. The ECB has included AI among the priorities of its banking supervision for the three-year period 2026-2028, with targeted workshops that photograph the vulnerability of banks in the face of a technology that can transform them from the inside before even threatening them from the outside.
On the credit front, regulators are reportedly asking individual institutions for details on their exposure to AI, including data center financing, amid growing concerns about the possible formation of a financial bubble in the sector.
From the problem of technological concentration we move on to market concentration and its effects on systemic risks: dependence on foreign technological giants, almost all of them American, risks exposing the continent’s banks to growing systemic threats. A bank that has financed a data center is exposed not only to the risk that that loan will not be repaid, but also to everything upstream: energy suppliers, local utilities, hardware manufacturers, generating a cascading risk that is difficult to capture with traditional models.
