GPT-5 AI model fuels German quantum cryptography breakthrough and telecom security testing

Two independent research teams solved an open problem in quantum cryptography using OpenAI's GPT-5 model. A MIT PhD student, Seyoon Ragavan, and professors Prabhanjan Ananth (UC Santa Barbara) and Amit Sahai (UCLA) posted their findings on arXiv within three hours of each other, demonstrating a solution for "unclonable encryption" that leverages quantum properties. They are now considering merging their work and note that the ease of accessing the same AI model raises questions about what counts as an original discovery. As Ananth put it, “If someone poses an open problem, the first thing you do is see whether GPT can solve it.”

Meanwhile, Deutsche Telekom is testing OpenAI’s GPT-5.5 Cyber to extend the reach of its existing security‑assessment processes. The company used the model to assist red‑team analysts in scanning internet‑exposed endpoints and to automatically review thousands of lines of source code in GitLab repositories, uncovering additional attack vectors and risky code patterns. CSO Thomas Tschersich emphasized that the goal is not to replace traditional security methods but to broaden their coverage without proportionally increasing resource use.