Philosophy
Why this exists
AICR embraces artificial intelligence to accelerate progress in cryptology. Its primary objective is to resolve outstanding conjectures and systematically formulate robust research programs. Given the sheer enormity of this undertaking, a deeply collaborative effort from the global cryptography community is essential.
Currently, AI cryptographic research often takes place in isolated, private notebooks. Without a common nexus to connect these efforts, efficiency is lost, efforts are duplicated, and the publication of results is delayed. AICR addresses these shortcomings by establishing an open, collaborative, transparent, and verified repository for AI-assisted research. The platform serves to aggregate and prioritize open questions. Token-expensive and often inegalitarian computational tasks can be distributed across the community. As AI advances and resources are allocated, formidable problems in cryptography gradually fall within reach.
The economics of problem-solving are becoming more explicit, with token expenditure and computational pricing emerging as dominant factors. In this landscape, Paul Erdős’s tradition of pricing mathematical conjectures is reimagined. Prizes can be offered transparently, including on-chain bounties payable upon the submission of verifiable zero-knowledge proofs, creating an open, decentralized market for cryptographic solutions. We acknowledge the longer-term effects of AI on academia and the hard philosophical problems it poses. However, at least in the coming years, the playing field is much more even, empowering a wider community to make meaningful progress.
Verification is critical, though we believe it must exist on a spectrum that carefully balances rapid progress with formal soundness. This dynamic demands acute community vigilance: autoformalization coupled with rigorous human review. In this evolving paradigm, we posit that reviewers are the future authors. Meticulous validation and prioritization will drive the field forward.
AICR thus provides a complement to traditional academic venues that, for now, impose limits on the use of generative AI in the publication process, from paper production to revision and acceptance. The best results found by the community can be selected for appearance in online proceedings.
This site has not been reviewed either: its prose, schema, scripts and this page were AI-generated, directed by a human but never independently reviewed, and the notice narrows only as review happens.
Our near-term focus is a dynamic, AI-native repository of open cryptographic problems — prioritizing a small, rigorously audited corpus over unvetted AI-generated volume. General-purpose archiving and other secondary functions are deliberately deferred for now, so the platform can support a hybrid solving model: external researchers drive the work, helped by structured problem packets, a formal record of past attempts, and occasional controlled model campaigns. This means a two-track submission pipeline — a formally vetted archive kept separate from an incubator for problems still being formulated.
Keeping this trustworthy calls for a fine-grained, versioned model that clearly separates a problem statement from an attempt, partial progress, and a claimed resolution. AI provenance is tracked precisely — from “AI-assisted” through to “AI-led” — so no contribution’s origin is left ambiguous, alongside verification methods that combine rather than replace each other. A machine-checked formalization or a proof-assistant kernel check is always paired with an independent human audit of whether the formal statement actually matches the informal one, holding the same standard of rigor the field expects of any result.
Staying sustainable long-term means taking risk and attribution seriously now: durable, citable credit across the different roles a piece of research actually involves, clear licensing (CC BY 4.0 for prose, MIT/Apache-2.0 for code), and a private channel for responsible disclosure. It also means resolving some real administrative exposure early — including a targeted trademark check on the project’s own name, given a potential conflict with existing marks — rather than treating that as someone else’s problem later.
We invite researchers to bring a problem, a proof, a refutation, tokens, or infrastructure to help build the repository. August 2026 marks an inflection point in cryptologic research; AICR attempts to bring the cryptology community together to build a new foundation that will shape future progress.