DeepMind's AI exposed old encryption flaws 1,600x faster — modern keys safe

Google DeepMind said on November 6 that its AI agents, working under human supervision, found two new ways to attack legacy 1024-bit RSA encryption — one that forges digital signatures and one that factors many keys at once, far cheaper than before.

The headline number is theoretical: for a short 32-byte fingerprint, a new algorithm improves on the previous best attack by about 1,600 times. The measured result is more modest — on an already-solved benchmark key, the agents' implementation reproduced the result with 3.5 times less CPU time. A second algorithm cuts the average cost of factoring each additional 1024-bit key by about 30 times.

The catch that keeps this from being a catastrophe: 1024-bit RSA is obsolete, and modern 2048-bit keys are unaffected. DeepMind's point is that AI can now compress cryptanalysis work that used to take years into weeks — so the math underneath today's encryption needs re-checking before attackers do it first.