On 30 April, I joined a panel at the Hertie School of Governance in Berlin to discuss how digital states continue to function during crises. Ukraine's Diia ecosystem (now serving over 23 million users) was our central case as we explored what genuine crisis preparedness means for Europe. Here are my key takeaways: Resilience is architectural, not an add-on. We need to look past the shiny surface of government apps to the underlying stack. Digital state resilience requires a r

Ungefähre Übersetzung des Beitrags "Cyber Defense Against AI-Enabled Attacks: Beyond the Mythos Hype" vom 24.04.2026 Die Debatte um Claude Mythos und ähnliche Frontier-Modelle wie Chat -GPT 5.5 hält an. Obwohl ihre genauen Fähigkeiten nicht öffentlich evaluierbar sind, könnten ihre fortschrittlichen Analysefähigkeiten als "Force Multiplier" für Cyber-Angreifer dienen, indem sie die Schwelle zum Finden von Schwachstellen und deren Ausnutzung mittels komplexer Exploits senken.

The noise around Claude Mythos and comparable frontier models has overshadowed a quieter truth: for most organisations, the most pressing cybersecurity challenges are structural rather than novel. At present, the most crucial defence lies in the consistent, thorough application of security fundamentals—done properly, at scale, together, and assisted by AI tools where appropriate. This matters most for the organisations least equipped to respond: resource-constrained municipal

The announcement of Claude Mythos has sent shockwaves through the cybersecurity ecosystem. Reactions to the latest frontier model are polarized: Debates range from whether the model marks the beginning of an existential crisis for cybersecurity or is merely an overhyped marketing stunt. Complicating matters, the model remains unavailable to the public. The reality is characteristically complex. While the dread surrounding Claude Mythos is partially fueled by a highly successf