Europe’s banks have a problem they can’t hack their way out of. Anthropic’s Claude Mythos Preview, the AI model capable of identifying and exploiting zero-day software vulnerabilities, remains available only to a handful of US partners. The EU has been trying to change that, and it’s not going well.
Talks between the European Union and Anthropic over testing arrangements for European financial institutions and companies have effectively stalled, according to reports from Spanish officials in mid-May 2026. That means the continent’s critical financial infrastructure still can’t stress-test itself against one of the most powerful cybersecurity tools ever built.
What Mythos does and who gets to use it
Anthropic released Claude Mythos Preview on April 7-8, 2026. The model demonstrated something that made cybersecurity professionals both excited and uneasy: advanced capabilities in finding and exploiting zero-day vulnerabilities, the kind of software flaws that exist before anyone knows about them, making them extraordinarily dangerous.
In English: Mythos can probe systems the way a sophisticated attacker would, but at a speed and scale that no human red team can match. For a bank or tech company, running Mythos against your own systems is like hiring the world’s best burglar to test your locks.
