Google DeepMind says its AI agents, working with human experts, found two weaknesses in legacy 1024-bit RSA encryption — in a matter of weeks. The announcement came on November 6 from Pushmeet Kohli, DeepMind's VP of Science.
Before the headlines scare you: only obsolete 1024-bit systems are affected. Modern 2048-bit keys — the ones behind banking, messaging and software updates — are not.
Two results, two papers
The work is split across two working papers, "Modular Roots from a Restricted Oracle" and "Fan Factories", led by DeepMind researchers Dj Dvijotham and Abhradeep Thakurta with cryptographers Nadia Heninger (UC San Diego) and Emmanuel Thomé (Inria). Humans directed the research; the models did much of the exploration. The papers are published in full but not yet peer reviewed.
Weakness one: forging signatures
The first finding is about digital signatures — how a computer checks that a message or software update really comes from its claimed source. The agents found that protection depends on how much of the signed block an attacker controls: once that share reaches roughly half, padding stops adding real protection, and forging a signature becomes almost as easy as if the padding were not there.
In a sandboxed test on an already-solved 1024-bit benchmark key, the agents' implementation reproduced the result with 3.5 times less CPU time. The eye-catching 1,600x figure is from a harder case — an attacker controlling only a small piece of signed data, like a 32-byte fingerprint — and it is a theoretical estimate from the maths, not a timed attack.
Weakness two: factoring in bulk
The second result targets factoring many RSA keys at once. The agents found a new algorithm with better asymptotic complexity than any previous batch method, cutting the average cost of factoring each extra 1024-bit key by roughly 30 times.
Again, that is an estimate, not a measured attack — nobody has built the "fan factory" the paper implies. But the direction is clear: cracking a big batch of old 1024-bit keys at once just got cheaper on paper.
Why 2048-bit stays safe
Every report of the announcement repeats this: the findings apply only to retired 1024-bit RSA. NIST disallowed 1024-bit RSA for generating new signatures back in 2013, which is exactly why researchers use it as a sandbox — complex enough to matter, but safe to attack without touching anything live.
The real story is the method. AI agents that can do genuine cryptanalysis in weeks give defenders a powerful new tool — the same kind of capability that recently let AI find hundreds of real Linux kernel flaws. The question is no longer whether AI can do security research; it is how fast defenders move once it does.
Read the full story on Rohan Tech Official: Google DeepMind's AI Agents Found Two Weaknesses in 1024-Bit RSA — 2048-Bit Keys Still Safe
Top comments (0)