Will an AI frontier lab crack a commonly used cryptography standard by end of 2026?
Delphi’s forecast
- Yesyour pick0.8%$0.01
- No99.3%$0.99
The forecaster strongly favors one outcome, so we can’t open a market for this question. You can still create this market yourself.
Forecast reasoning
Frontier AI labs have found novel attacks on non-deployed schemes (HAWK, reduced-round AES) and traced NIST curve origins, but as of October 2026 no commonly used cryptography standard has been cracked. The 74-day window to resolution is tight, and no evidence suggests imminent breaks of deployed standards like RSA or AES-256.
Settlement
Resolution criteria: - "Yes": Wins if, strictly after 2026-10-08T17:53:17.348Z and at or before 2027-01-01T00:00:00.000Z, at least one AI frontier lab (e.g., OpenAI, Anthropic, Google DeepMind, Meta AI, xAI, or a comparable organization at the leading edge of AI capability) publicly demonstrates a practical break of a commonly used cryptography standard as defined in Clarifications — meaning a complete key recovery, full decryption of ciphertext, or equivalent total break against the full-strength, production-parameter version of the standard, not a reduced-round variant, a challenge benchmark, or an undeployed candidate. - "No": Wins if no such demonstration occurs strictly after 2026-10-08T17:53:17.348Z and at or before 2027-01-01T00:00:00.000Z. Clarifications: - "Commonly used cryptography standard" means an algorithm actively deployed at scale in production systems as of the market's open, including but not limited to: full AES-128, AES-256 (all rounds, production parameters), RSA-2048 or larger (as used in TLS/PKI), ECDSA/ECDH on NIST P-256 or P-384, SHA-256, SHA-3, and ChaCha20-Poly1305. Excluded are: NIST post-quantum candidates not yet standardized and not yet widely deployed (e.g., HAWK), RSA challenge benchmarks below 2048-bit key length (e.g., RSA-896), reduced-round or otherwise weakened research variants of any algorithm, and historical or academic ciphers not in active production use. - A "practical break" requires that the attack succeeds against the full unmodified standard at real-world security parameters, not merely an improvement in theoretical attack complexity that still leaves the standard computationally secure in practice.
Jan 1, 2027, 12:00 AM UTC