Google DeepMind’s AI agents have made a significant breakthrough in cryptography, finding two ways to reduce the cost of attacking 1024-bit RSA cryptosystems. This has major implications for legacy systems that still use these outdated keys, but modern encryption protocols remain secure.
According to Pushmeet Kohli, VP at Google DeepMind and chief scientist at Google Cloud, the research conducted with cryptographers from UC San Diego and Inria indicates that attacks on 1024-bit RSA require less computation than previously assumed. This is significant because it means that certain types of attacks can be executed more quickly and with fewer resources.
The first finding involves digital signatures and padding, which are added to a message to bring it to a fixed block size. It was known that padding made forgery harder, but its value against an attacker who can choose what gets signed was unclear. DeepMind determined that the effectiveness of padding depends on how much of the block the attacker controls. If they control roughly half or more of the block, padding adds no protection at all. This applies to the PKCS#1 v1.5 format used by the PKCS#11 signing interface of cryptographic devices such as hardware security modules (HSMs).
In a sandboxed test, DeepMind’s AI agents re-ran an earlier academic forgery attack on the same 1024-bit key, cutting CPU time 3.5-fold. The first forgery took 391 CPU core-years, whereas in the original study it took 1,380. Two AI agents designed, ran and debugged the forgery computation in under four days, with approximately 20 messages from a human researcher. The key was not factored; that would take an estimated 500,000 to 1 million core-years.
The second finding concerns factoring many RSA keys of the same size, where an attacker can share the cost of one large upfront computation across all of them. DeepMind’s agents came up with an algorithm that makes factoring each additional 1024-bit key roughly 30 times cheaper on average than the best previous approach. However, this theoretical estimate assumes an initial computation as costly as factoring one key.
It’s essential to note that these results only affect obsolete 1024-bit RSA systems and do not impact modern encryption protocols, which use keys of 2048 bits or larger. Kohli emphasized that using AI to rigorously test previously held assumptions will ensure that modern encryption stays secure in an era where AI-enabled mathematics and computation are becoming increasingly prevalent.
In practical terms, this breakthrough highlights the importance of regularly reviewing and updating cryptographic systems. Organizations with legacy RSA-1024 systems should consider migrating to more secure protocols as soon as possible. While factoring a single key is still a significant computational task, sharing the cost across multiple keys makes it more feasible for attackers. As AI continues to advance and become more accessible, it’s crucial that we stay ahead of potential threats by regularly reassessing our cryptographic defenses.
Source: SecurityWeek — 2026-11-09