[REVISION HISTORY]

AI advancements in cryptographic problem solving

Updated 1 time since CLSTR started tracking revisions of this situation.

What changed

2026-10-17 15:12 UTC → 2026-10-18 12:22 UTC · added ~~removed~~

Recent developments indicate that AI models, specifically Anthropic’s Claude, are increasingly capable of solving complex cryptographic challenges. In September 2026, researcher Steve Weis used Claude AI to factor the RSA-896 semiprime, a 270-decimal-digit number. By utilizing the General Number Field Sieve (GNFS) algorithm optimized through AI to manage GPU computations, Weis set a new public factoring record. While experts state this does not yet compromise the RSA-2048 standard, it suggests that RSA-1024 keys may be vulnerable to entities with sufficient GPU resources. By October 2026, Claude AI was used to resolve a 25-year-old mystery concerning the origin of cryptographic seeds used in several NIST-standardized elliptic curves. The AI identified in under 30 minutes that the seeds used by the NSA for curves such as P-256 and P-384 were generated via the lrand48 pseudo-random number generator from a 1990s C library. ~~This finding suggests~~ Research released on October 16, 2026, specifies that the seeds ~~resulted from an ordinary computational process rather than deliberate~~ for curves P-192, P-224, P-256, P-384, and P-521 were initialized with Unix timestamps between May 1997 and April 1998. The model recognized a pattern where lrand48 outputs only 31 bits, resulting in systematic zero bits in the seeds. While the discovery does not expose immediate security vulnerabilities or provide a method to break existing encryption, it provides reproducible evidence that reduces long-standing suspicions of intentional mathematical backdoors. Anthropic donated the associated $36,864 bounty to the Internet Security Research Group.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.