The European Central Bank has ordered major euro-area banks to prepare for a new breed of cyber threats. The deadline: end of October. This isn’t a drill—frontier AI models are rewriting the cyber risk playbook.
Claudia Buch, chair of the ECB’s supervisory board, spelled it out bluntly: advanced AI can now uncover hidden software flaws and craft working exploits at unprecedented speed. The ECB demands banks patch software faster, strengthen AI-powered defenses, and tighten control over third-party tech providers.
The ECB isn’t just asking nicely. It plans to rank banks by their readiness and pressure laggards to catch up. They’ve already run 109 banks through a severe cyber-attack drill. The message: get serious or fall behind.
At the heart of this threat is Mythos, an AI model from the US startup Anthropic. Launched initially in April 2026 and updated in June, Mythos can find and weaponize vulnerabilities in major operating systems and browsers—some unknown for decades. The US government called it a “reckoning” and barred all users from access, restricting Mythos to trusted American organizations.
Mythos isn’t an isolated case. Sysdig researchers recently documented the first autonomous AI ransomware attack. An AI agent named Jadepuffer breached a vulnerable server, stole login credentials, encrypted a production database, and demanded Bitcoin ransom—all without human intervention.
The stakes are systemic. The European Systemic Risk Board raised the cyber risk level to “severe,” citing frontier AI as a systemic threat. Their warning cuts through the noise: “The timeline is not years, it is months.” Cyber risk is no longer just a technical problem. It’s a core business risk demanding leadership at the highest level.
Europe’s challenge deepens as nearly all leading AI providers operate outside its borders. This dependence leaves the EU vulnerable. French startup Mistral is trying to break the mold, offering flaw-hunting tools and aiming to design proprietary AI chips to reduce reliance on Nvidia. Still, Mistral controls only a fraction of the US’s 25 gigawatts of AI compute power.
The geopolitical angle sharpens the picture. The Five Eyes intelligence alliance—Australia, Canada, New Zealand, the UK, and the US—issued a joint alert emphasizing that frontier AI models pose the greatest risk to critical systems within months. China claims to have developed an AI “cyber nuclear weapon” rivaling US models, heightening tensions in cyber warfare.
AI-powered cyber attacks could target power grids, financial networks, satellite communications, military systems, and even nuclear facilities. The damage could cascade, causing power outages, logistics collapses, and disruptions to hospitals and airports.
So far, no catastrophic AI-driven cyber attacks have hit. But the warnings are urgent. ECB President Christine Lagarde cautioned AI might trigger a dangerous financial crisis. The combined message from regulators and intelligence agencies is clear: the window to protect critical infrastructure is closing fast.
In the end, Europe risks becoming an economic backwater if it fails to build its own AI capabilities. The technology is evolving at breakneck speed. Banks, governments, and startups must act now or pay dearly later.
Based on
- ECB tells Europe’s big banks to plan for AI cyber threats — thenextweb.com
- AI ‘cyber weapons’ could target America’s power grid, experts warn | The Independent — independent.co.uk
- What an AI ‘cyber nuclear war’ would actually look like | The Independent — independent.co.uk
- Warning issued as Bitcoin cyber attack is carried out fully by AI | The Independent — independent.co.uk
- Europe should build AI – and fear it, too | Semafor — semafor.com
