Anthropic publishes a practical key-recovery attack on HAWK-256

Who Should Read

A backend developer responsible for crypto choices and security design who wants to follow recent cryptanalysis research

Core Mechanics

• A cryptanalysis code repo is out
  • It is the GitHub repo anthropics/cryptography-research-demo.
  • The README says it is cryptanalysis code accompanying associated papers.
• Three folders: AES, HAWK, LEA
  • The three components are independent.
  • They live in AES/, HAWK/, and LEA/.
• Research artifact, not maintained
  • The README says Research artifact, Not maintained and not accepting contributions.
  • It is licensed under Apache-2.0.
• The README gives no attack details
  • No numbers, attack method, or complexity are in the source.
  • The repo has 2 commits, 70 stars, and 8 forks.

Community Reactions

• A commenter said they believed this class was already broken
  • They still found it cool because the attack is practical.

How to Apply

• List where you use AES, LEA, or HAWK
  • Scan your services today and tabulate uses of these three algorithms.
  • If you use HAWK anywhere, check the HAWK/ folder and its linked paper.
• Schedule a review of HAWK prototypes
  • If you use HAWK even experimentally, read the paper with your security owner.
  • Check whether the attack applies to your parameters.

Terminology

Related Papers

• The Judge Can Debias Itself: Awareness and Self-Debiasing for Reliable LLM-as-a-Judge

  Warning an LLM judge about bias or letting it self-critique improved 20 of 21 model–bias combinations.

• 90% of US working hours are spent on tasks not covered well by AI benchmarks

  Mapping 29 benchmarks to job tasks shows 90% of US working hours lack good coverage

• Postgres 19 delayed after AI reviewers find more bugs than humans can fix

  LLM code review surfaced so many bugs that Postgres 19 slipped to November and 8-10 features were reverted

• Show HN: Gemini 4 Argon plays StarCraft

  LLMs wrote StarCraft bots in one hour; GPT-6 Astra scored highest.

• A Plausible Explanation Is Enough to Mislead Multimodal Judges

  Adding plausible explanations to fixed answers cut MLLM judges' wrong-answer rejection by 12.7–13.8 points.

• Neural networks are clean

  Looking at neurons directly shows most have one main job, challenging superposition.

Related Resources

• https://github.com/anthropics/cryptography-research-demo
• https://news.ycombinator.com/item?id=49087091