bitcoin ·
OpenAI claims a cryptography breakthrough. Does that threaten Bitcoin?
The claim by OpenAI regarding a cryptographic breakthrough, which touches upon the Exponential Time Hypothesis (ETH), is currently subject to independent review, and a practical assault on Bitcoin…
The claim by OpenAI regarding a cryptographic breakthrough, which touches upon the Exponential Time Hypothesis (ETH), is currently subject to independent review, and a practical assault on Bitcoin would necessitate several subsequent developments.
The announcement made on Oct. 26 stated that OpenAI's artificial intelligence system managed to disprove the Exponential Time Hypothesis, which has long been a foundational belief concerning the inherent difficulty of solving certain logical problems. Researchers are now arranging an independent investigation into this finding. It is important to note that the information disclosed does not prove a practical vulnerability against Bitcoin. The ETH referenced here stands for Exponential Time Hypothesis, not Ether. The specific area of focus is 3-SAT, which involves a computational task where a machine must locate true-or-false assignments that satisfy a set of logical constraints. The hypothesis posits that solving these problems requires an amount of time that grows exponentially relative to the quantity of variables in the most challenging scenario.
OpenAI asserted that the outcome was validated within Lean, a type of proof assistant, although the scope outlined in its notes suggests a less stringent time limit compared to the most definitive assertion made in the manuscript. The group of researchers coordinating the community's review has explicitly stated that their participation should not be interpreted as an endorsement of the claim. Their correspondence noted, "Our involvement should not be taken as certification of the claim." The document, which was dated Oct. 17 and subsequently revised on Oct. 25, puts forth a deterministic method capable of precisely counting the assignments that satisfy fixed-width Boolean formulas in subexponential time while using polynomial space. Put simply, the time required for this process increases at a slower pace than an exponential rate as the problem size increases. This represents a different promise than rendering the problem tractable on contemporary computing hardware. Furthermore, the disclosed boundary does not establish an algorithm that runs in polynomial time, nor does it resolve the P=NP question, which asks whether every decision problem whose proposed solutions can be verified in polynomial time can also be solved within polynomial time.
CryptoSlate's analysis of the paper is confined to its opening pages, pages 1–8, rather than encompassing a full review of the entire proof. The ETH refutation claimed by OpenAI pertains specifically to formula solving, whereas mining and spending represent different, concrete operational duties. Dashed connectors indicate a practical consequence that has not been established, rather than a pathway for an attack.
A theoretical speed-up in SAT does not, on its own, demonstrate an advantage in mining. Searching for a block-header hash is the function of mining, while signatures grant permission to spend. The SAT claim does not provide evidence of retrieving a secp256k1 private key. A subexponential SAT bound alone cannot prove a practical Bitcoin attack. The ETH refutation is OpenAI's assertion pending this independent scrutiny. The paper's evidence is restricted to the introductory pages 1–8; there are no established attack times, encoding sizes, performance benchmarks, or independent cryptographic audits.
Bitcoin utilizes ECDSA and Schnorr signatures over the secp256k1 elliptic curve, as detailed in BIP340. The security assumptions mentioned there relate to hash functions and elliptic-curve discrete logarithms, and BIP340 also notes stronger assumptions for ECDSA security proofs. Bitcoin mining involves searching for a block-header hash that falls beneath a specified target, while transaction signatures serve to authorize expenditures; these are distinct cryptographic functions. The theoretical SAT speed-up that was revealed does not inherently prove a practical Bitcoin attack; implementing it would require a concrete encoding and a comparison of costs for the specific cryptographic task involved.
Bitcoin's mechanism for adjusting difficulty recalculates its mining target every 2,016 blocks, but this process is separate from the authorization provided by signatures. OpenAI maintains that translating these theoretical advancements into functional improvements demands considerable additional research.
Ryan Williams, posting under the handle @rrwilliams, mentioned that he, along with Russell Impagliazzo and Mohan Paturi, was asked to organize an open, independent examination without receiving NDAs or payment. In a separate communication, @rrwilliams shared that he had requested OpenAI to verify if existing cryptographic systems were compromised, relaying that they appeared secure. This is a secondary report of a check requested from OpenAI, not an independent cryptographic audit. OpenAI itself states that achieving practical enhancements requires substantial further investigation. When asked about potential benefits for developers, @rrwilliams stated he was unaware, leaving open the possibility for either progress within 12 months or no such development. The organizers plan to disseminate their findings as the examination continues.
Liam Wright, who is a reporter, podcast producer, and Editor-in-Chief at CryptoSlate, is also known as "Akiba," and he believes that decentralized technology possesses the capacity to facilitate...
The key-recovery question centers on whether an adversary can deduce a private key from its corresponding public key. Disproving an assumption about worst-case SAT solving does not, by itself, demonstrate a method to overcome these specific mechanisms. A functional attack would still necessitate a suitable encoding, an implementation, and a cost comparison tailored to the targeted task. Furthermore, the variable count within a SAT formula cannot be directly equated with the bit length of a cryptographic key or hash. The Lean scope measures its bound based on the variables and input length of the encoded formula. Applying a standard crypto bit count to that bound would bypass the necessary work of representing the actual problem. Bitcoin's difficulty adjustment recalculates its mining target every 2,016 blocks, a function pertaining to block production, not signature authorization, and this would not render a private mining advantage moot.
This distinction follows a discussion CryptoSlate covered on Oct. 7, during which Justin Drake called for preparations for "bunker mode" following an earlier mathematics release from OpenAI. That earlier warning concerned a potential future attack, and that initial announcement did not demonstrate a practical break of ECDSA or RSA. The ETH claim from Oct. 26 addresses a different outcome.
Earlier coverage
Reported by cryptoslate.com.
