GPT-5.6 Sol Ultra Drives 2 Proofs of Quantum Encryption Breakthrough in 1 Day
Updated
Updated · Scientific American · Jul 31
GPT-5.6 Sol Ultra Drives 2 Proofs of Quantum Encryption Breakthrough in 1 Day
2 articles · Updated · Scientific American · Jul 31
Summary
Two arXiv preprints submitted last Thursday by separate teams claimed efficient unclonable encryption without prior security caveats, and both credited GPT-5.6 Sol Ultra with generating the core proof ideas.
10:35 a.m. PDT marked the first submission from Prabhanjan Ananth and Amit Sahai; MIT student Seyoon Ragavan filed a second 3 hours 18 minutes later after using the same model through a different workflow.
The result targets a niche quantum cryptography problem first framed in 2019: encrypting a message so a stolen quantum ciphertext cannot be split into two usable copies once a key is known.
Researchers say the papers still need peer review, but Anne Broadbent and Andrea Coladangelo called the arguments convincing even as they warned AI could compress independent discovery and displace graduate-student work.
The near-simultaneous proofs highlight a broader shift in theoretical computer science, where open problems can now be attacked immediately with shared AI tools, complicating credit, training and access.