Google DeepMind says its AI agents, guided by human cryptographers, have found two weaknesses in legacy 1024-bit RSA encryption, work that would normally have taken researchers far longer. The findings were announced on 6 November by Pushmeet Kohli, DeepMind's vice president of science.

The first weakness concerns digital signatures, the mechanism that confirms a software update, document or website really comes from its claimed source. Signing schemes add padding, extra preset data meant to make forgery harder. DeepMind's agents found that once an attacker controls roughly half of the signed data block or more, the padding adds virtually no extra protection. In a sandbox test against an already-solved 1024-bit benchmark key, the agents' code reproduced the forgery using 3.5 times less CPU time than the previous method.

The second weakness targets batch factoring, breaking many keys of the same size at once. DeepMind's working paper, titled "Fan Factories" and dated 5 November, describes a new algorithm that it estimates cuts the average cost of factoring each additional 1024-bit key by about 30 times compared with the best earlier batch method.

The research was led at DeepMind by Krishnamurthy (Dj) Dvijotham and Abhradeep Thakurta, working with cryptographers Nadia Heninger of UC San Diego and Emmanuel Thomé of Inria. DeepMind stresses that the work was AI-assisted research, not autonomous discovery, with human experts directing the agents throughout.

Both results apply only to obsolete 1024-bit RSA. Modern systems almost universally use 2048-bit keys or larger, and DeepMind says those are unaffected, so online banking, messaging apps and software updates are not threatened, and no live keys were broken. The larger speedup figures are theoretical estimates, not measured attack times. It still matters because retired does not mean gone: the US National Institute of Standards and Technology has disallowed 1024-bit RSA for new signatures since 2013, yet such keys persist in older hardware, long-lived certificates and embedded systems.