Matthew Green's AI-run study quantifies cryptanalytic effort: lattices now 1.7x ECDLP
matthew_d_green · x · 2026-10-14
Cryptographer Matthew Green oversaw a study built and executed largely by AI models (Claude Fable 5.1, Claude Opus 5.5, with GPT-6 Astra), using the literature itself to estimate public cryptanalytic effort behind public-key assumptions.
- 16,000 candidate papers were classified for work on ECDLP, lattices (MLWE/LWE), and syndrome decoding, converted into research-allocation years via an author-level map.
- Key findings: lattices only recently overtook elliptic curves in cumulative effort, now at 1.7x ECDLP; 269x work remains beyond the largest public P-256 solve; corpus spans 2,785 works and 3,240 distinct authors.
- The authors caution publication-based estimates may substantially overstate actual time spent.
- Green used the study to argue to Bram Cohen that factoring, FFDLP, ECDLP, and MLWE are all subject to potential increases in attack efficiency. Working draft, open for corrections.
More from Safety
- Super-Worms Warning: AI-Powered Self-Propagating Malware and a Fragmenting Internet — JensHonack · 2026-10-23
- Claude Code Pushes Urgent Security Update Across All Installations — umeshonai · 2026-10-23
- Sanders calls to pause autonomous AI agents in bank accounts; critics push back — AIandDesign · 2026-10-23
- TechPolicy essay: system prompts are weak evidence of AI safety, outputs must be evaluated — mmitchell_ai · 2026-10-23
- Anthropic's deputy security chief warns cyber threats will get 'a lot worse before they get better' — Polymarket · 2026-10-23
- AI cyberattacks could paradoxically accelerate AI development in an offense-defense arms race — kimmonismus · 2026-10-23