Abstract glowing qubit spheres linked by entanglement threads
AI-generated image

Algorithms

DeepMind says AI agents uncovered two weaknesses in legacy 1024-bit RSA

Published · Last updated

Short answer: Google DeepMind announced on X on 6 November 2026 that human-guided AI agents uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems. A follow-up post said modern 2048-bit systems remain safe from those specific exploits. Demis Hassabis amplified the thread.

DeepMind framed the work as accelerating the science of stress-testing encryption. VP of science Pushmeet Kohli's linked essay, cited in the DeepMind posts, expands on signature and factorization advances against obsolete 1024-bit schemes. DeepMind's own caveat is that historical hardness assumptions need continuous testing as AI-enabled math improves.

What defenders should take away

The public claim is about legacy RSA-1024, not a break of current recommended key sizes. DeepMind presents the results as a reason to keep stress-testing older standards ahead of future threats. Until peer-reviewed writeups and independent reproductions land, treat the DeepMind X announcement as the company's research notice.

← All news