EN ▾
SNAPE REP RT Front page →
ConfirmedAI safety06 Nov, 17:20 UTC

DeepMind AI agents uncover weaknesses in legacy 1024-bit RSA systems

Google DeepMind mark for cryptanalysis research
Google DeepMind

Google DeepMind 17:20:31Z: human-guided agents found signature-padding and batch-factorization advances on obsolete RSA-1024; modern 2048-bit+ keys remain secure.

Read at Google DeepMind Share

Why it matters

DeepMind’s Science team says AI agents, steered by humans, accelerated cryptanalysis against retired 1024-bit RSA proving grounds: one result weakens padding’s protection once an attacker controls roughly half a signed block, another cuts average cost to factor additional same-size keys in a batch by about 30× versus prior methods. The lab stresses modern 2048-bit and larger keys stay out of scope. The point for singularity watchers is tempo: hardness assumptions can now be stress-tested faster than the old multi-year cadence.

Key facts

  • Posted 17:20:31Z Nov 6
  • Targets obsolete RSA-1024 only
  • Signature-padding result
  • ~30× batch factorization gain claimed
  • 2048-bit+ keys remain secure

Sources

More on AI safety
More from Snape ReportFront page →