DeepMind says human-guided AI agents found two weaknesses in legacy 1024-bit RSA
Google DeepMind announced research in which its agents, guided by researchers including Nadia Heninger and Emmanuel Thomé, found two weaknesses in 1024-bit RSA. OfficeChai reports one concerns signature padding, with a theoretical gain of about 1,600 times in one setting and a 3.5 times CPU saving reproduced on an already-solved benchmark key. The other is a batch method said to cut the average cost of factoring each additional 1024-bit key by about 30 times. DeepMind says modern keys of 2048 bits or more are not affected.
Why it matters: It is another sign that AI agents can speed up cryptanalysis, which matters for anyone still running old key sizes.
| Benchmark | Score | Type | Measured or published by |
|---|---|---|---|
| Signature padding attack, theoretical gain | About 1,600x | Cited | Google DeepMind, via OfficeChai |
| Batch factoring, cost per additional 1024-bit key | About 30x lower | Cited | Google DeepMind, via OfficeChai |
We could not read DeepMind's own post or a paper; its X thread blocked automated reading. The models used were not named.