The debate about the risks of current AI-based Large Language Models (LLMs) for cybersecurity has gained significant momentum following the release of Anthropic’s latest model, Claude Mythos. Initially withheld from publication (presumably due to fears of misuse), then released with safety precautions, only to be placed under a U.S. government export ban and taken offline shortly after—the saga of Mythos exemplifies the critical role of LLMs in cybersecurity. These cutting-edge AI tools threaten to scale, accelerate, and personalize existing attack methods, giving adversaries a significant advantage while putting defenders under immense pressure.
Bibliographic record:
Schörnig, Niklas. “Die dunkle Seite neuer KI-Sprachmodelle. Wie Large Language Models die Cybersicherheit herausfordern.” PRIF Spotlight 8/2026, 25.06.2026
