A Korean AI startup has run quantum cryptanalysis at unprecedented scale on live 156-qubit hardware — and published open tools so anyone can replicate the findings.
TL;DR: VIDRAFT, a Korean Pre-AGI AI startup, has successfully demonstrated quantum cryptanalysis of two foundational symmetric-key cipher structures — Even-Mansour and a reduced Feistel scheme — on IBM's 156-qubit quantum computer. Using Simon's algorithm on real hardware rather than a simulator, the company recovered hidden periods embedded in each cipher structure and applied a self-verification procedure to confirm the results were not pre-engineered. VIDRAFT stresses this is a proof-of-concept, not a break of AES or any production cipher.
VIDRAFT, the Korean Pre-AGI AI startup behind the Darwin language model family and the AETHER architecture, announced on July 6, 2026 that it has successfully conducted quantum cryptanalysis experiments on an IBM 156-qubit quantum processor, targeting reduced versions of two widely studied symmetric-key cipher structures: Even-Mansour and Feistel.
The core of VIDRAFT's experiment was the recovery of hidden periods — mathematical structures embedded inside cipher designs that, if exposed, can reveal information about secret keys. Working on IBM's live quantum hardware rather than a classical simulator, the team applied Simon's algorithm to probe these structures at a scale the company describes as unprecedented for noisy, real-world quantum devices.
For the Even-Mansour structure, VIDRAFT successfully recovered the hidden period underpinning the secret key across a range of input sizes. For a three-round reduced Feistel structure — a skeleton of the kind of design used in classical block ciphers — the team confirmed hidden period recovery under multiple block-size conditions.
Critically, VIDRAFT built a self-verification step into every experiment instance. Rather than simply claiming a correct result, the procedure independently recovers a second key value alongside the first, demonstrating that the algorithm genuinely found the hidden structure from the quantum output — not from prior knowledge of the answer.
The company was explicit about what the experiment does not claim: it does not represent a break of AES, DES, or any cipher protecting real financial or communications infrastructure. The Feistel target was a reduced, academic skeleton, not a full production cipher. VIDRAFT also noted that the work used noise-mitigation techniques rather than full quantum error correction, and frames the significance as running a cryptanalytic quantum algorithm at an unusually large scale on genuinely noisy hardware — not as proof that quantum computers have surpassed classical systems in a general sense.
Alongside the experimental results, VIDRAFT publicly released a suite of quantum cryptanalysis tools covering five representative cipher structures: linear ciphers, block ciphers and substitution-permutation networks (SPN), Even-Mansour, CBC-MAC, and Feistel. The tools are accessible directly from a web browser, letting researchers and security professionals explore how each structure could be affected by quantum algorithms without needing their own quantum hardware access.
VIDRAFT CEO Kim Min-sik commented that quantum computers, once mature, will become a pivotal tool for accelerating solutions to hard scientific problems in areas like drug discovery, new materials, and cybersecurity. He added that VIDRAFT intends to continue combining AI and quantum computing in empirical research.
The company plans to pursue academic publication and independent external verification of the results.
The cryptographic community has long known, in theory, that Simon's algorithm poses a structural threat to certain symmetric-key designs. What has been missing is empirical evidence at meaningful scale on real, error-prone quantum hardware — as opposed to small-scale demonstrations or purely classical simulations. VIDRAFT's work begins to close that gap.
For the security industry, this is an important signal rather than an alarm. The experiment confirms that quantum cryptanalytic algorithms can be executed on today's hardware in ways that are structurally meaningful, even if full-scale attacks on production ciphers remain far beyond current capabilities. Organizations assessing post-quantum migration timelines now have additional, concrete data to factor into those decisions.
VIDRAFT's decision to open-source the cryptanalysis tools further raises the bar for transparency: rather than presenting results that only the company can reproduce, they are inviting the broader research community to probe and challenge the findings. The planned paper submission signals a commitment to peer review that will determine how the security community ultimately rates the significance of this milestone.
Q: Has VIDRAFT broken AES or any real-world encryption?
A: No. VIDRAFT explicitly states that this experiment targeted reduced, academic cipher structures for proof-of-concept purposes and does not constitute a break of AES, DES, or any encryption protecting production systems.
Q: What quantum hardware did VIDRAFT use for this experiment?
A: VIDRAFT conducted the experiments on IBM's 156-qubit quantum computer, using noise-mitigation techniques rather than full quantum error correction.
Q: What cryptanalysis tools has VIDRAFT made publicly available?
A: VIDRAFT released web browser-accessible tools for quantum cryptanalysis of five cipher structure types: linear ciphers, block ciphers/SPN, Even-Mansour, CBC-MAC, and Feistel structures.
Source: 전자신문 (2026-07-06) — original article