Public story · 2026-07-30 · high

AI Hits 86% on Solved Ciphers, Under 9% on Unsolved

Anthropic built the benchmark with partners in Switzerland, Germany, and Israel; the models still found a real SpoC AEAD attack.

Why now: The tier breakdown is dated July 30, capturing where five frontier models stand on cryptanalysis right now.

Issue

2026-07-30

Confidence

high

Sources

Redaction

passed

Story

Anthropic and four university partners built a 191-task cryptanalysis benchmark, and frontier models topped out at 86% only on ciphers that are already broken.

That gap matters for anyone taking AI capability claims at face value. On the 142 schemes with no public break, every model scored under 9% at full strength, per the benchmark site.

The test spans six primitive families, drawn mostly from AES, SHA-3, Lightweight Cryptography, and the Post-Quantum NIST competitions. Tier 1 holds 49 already-broken schemes. Tier 2 holds the 142 that are still standing.

Five frontier models ran against both tiers. Claude Mythos 5 topped Tier 1 at 85.7%, the strongest of the five, according to the benchmark site.

The models weren't empty-handed on the hard tier, though. They produced a genuine key-recovery attack on the full SpoC AEAD cipher and caught an error in the published CCA-security proof for KINDI, per the benchmark's authors.

The 86% headline score is recall of public write-ups, not cryptanalysis skill. The real number is the under-9% score against ciphers nobody's broken yet.

The tier breakdown is dated July 30, capturing where five frontier models stand on cryptanalysis right now.

Source trail

Anthropic's benchmark with ETH Zurich, Tel Aviv University, University of Haifa, and TU Berlin

Entities

AES Anthropic CCA Claude Mythos CryptanalysisBench ETH Zurich Haifa KINDI Lightweight Cryptography POST Quantum NIST SHA SpoC AEAD Tel Aviv University Tier TU Berlin

Claim evidence

AI Hits 86% on Solved Ciphers, Under 9% on Unsolved
Source

Provenance

Canonical issue: 2026-07-30
AI generated: yes
Story unit: 2026-07-30-cryptanalysisbench-65-86-on-already-broken-ciphers-under-9-on-anything-without-a-published-break
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

AI Hits 86% on Solved Ciphers, Under 9% on Unsolved

Story

Related stories

AI Models Break a Cipher, Expose a Proof Error

Anthropic's Claude Mythos cut a HAWK post-quantum attack from 2^64 to 2^38 in about 60 hours.

Green confirms Anthropic's HAWK break, downgrades AES claim

Fable 5's filters are strict because it's built on a model Anthropic refused to ship.

Commerce Restores Claude Mythos 5 Access for About 100 Groups

Anthropic's next model is in confirmed preview.

Anthropic published the actual diagnostic for model tier versus effort level.

Anthropic, Atlassian, and a Scale VP All Landed on the Same Answer: Build the Skills Layer, Not a New Stack

Source trail

Entities

Claim evidence

Provenance

AI Hits 86% on Solved Ciphers, Under 9% on Unsolved

Story

Related stories

AI Models Break a Cipher, Expose a Proof Error

Anthropic's Claude Mythos cut a HAWK post-quantum attack from 2^64 to 2^38 in about 60 hours.

Green confirms Anthropic's HAWK break, downgrades AES claim

Fable 5's filters are strict because it's built on a model Anthropic refused to ship.

Commerce Restores Claude Mythos 5 Access for About 100 Groups

Anthropic's next model is in confirmed preview.

Anthropic published the actual diagnostic for model tier versus effort level.

Anthropic, Atlassian, and a Scale VP All Landed on the Same Answer: Build the Skills Layer, Not a New Stack

Source trail

Entities

Claim evidence

Provenance