An Open Quantum Encryption Problem, Solved Twice in One Day

Two teams, working separately from the same talk and the same AI model, posted proofs of the same result three hours apart. UCSB's Prabhanjan Ananth co-authored the first, and one of his doctoral students discovered the overlap.

Prabhanjan Ananth (UCSB) and Amit Sahai (UCLA) posted their proof to arXiv on July 23 at 10:35 a.m. PDT. Seyoon Ragavan (MIT) posted his three hours and eighteen minutes later. The two papers establish the same result. Both credit the same AI model with finding the idea at the center of it.

The result concerns unclonable encryption, a corner of quantum cryptography that exploits a property classical information does not have. An unknown quantum state cannot be perfectly copied. Encode a message into qubits and an adversary who intercepts the ciphertext should not be able to split it into two versions that both become readable once the decryption key is released. Anne Broadbent of the University of Ottawa and her then student Sébastien Lord introduced the modern framework for this in 2019. Earlier approaches were either inefficient or required assumptions to be secure. What remained open was whether an efficient scheme could deliver strong security with no assumptions at all.

Both new papers claim it can, for a single bit sent one time. The scheme itself is close to one Broadbent and several collaborators published earlier this year. The contribution is the proof that it offers the stronger security researchers had been seeking.

Ananth, an Associate Professor of Computer Science at UCSB, and Sahai, who directs UCLA's Center for Encrypted Functionalities, worked through a system built at UCLA to help AI models pursue and critique possible solutions. Their paper states that the model produced the construction and main proof ideas, and that the researchers refined and verified the work and take responsibility for its claims. Ragavan, a third-year doctoral student, took a hands-on route, directing the model in two-hour stretches and redirecting it between rounds.

Neither team knew of the other. Both had heard the problem raised at a talk at the Simons Institute for the Theory of Computing at UC Berkeley earlier that month. The overlap surfaced through UC Santa Barbara. Yao-Ting Lin, a doctoral student advised by Ananth, received a draft from Ragavan on the same afternoon he sat in a meeting where he heard about the second proof. He connected the two groups. 

Neither paper has been peer reviewed, and the two teams have since discussed combining them into a single version for possible conference submission. Broadbent and quantum computing researcher Andrea Coladangelo have said the results appear convincing. Broadbent has also raised concerns about the effect on graduate students, noting that work of this kind is what she would normally assign them, a concern Ananth shares.

Scientific American reported on the convergence on July 31: https://www.scientificamerican.com/article/ai-helped-produce-two-proofs-for-the-same-cryptography-problem/

Preprint: https://eprint.iacr.org/2026/1511