DeepMind RSA-1024 Research Finds Attacks 1600 Times Faster

DeepMind RSA-1024 research found faster ways to probe weaknesses in the older 1024-bit version of RSA encryption, while leaving modern 2048-bit and larger keys untouched. Interesting Engineering reported on November 6, 2026 that Google DeepMind's human-supervised AI agents identified techniques that can expose flaws in that legacy system much faster than earlier methods.

CryptoBriefing, writing the same day, said DeepMind announced two weaknesses in legacy 1024-bit RSA on November 6, 2026, and stressed that current deployments using stronger keys remain secure against this specific threat.

What did the DeepMind RSA-1024 research find?

The first result involves digital signatures, which computers use to check that software updates, websites, and documents come from a trusted source. Those schemes add extra preset data, known as padding, before a message is signed. The agents found that once an attacker controls roughly half of the signed data block or more, that padding offers virtually no extra protection, making forgery no harder than attacking an unpadded system.

Even when the attacker controls only a small slice of the signed data, such as a 32-byte digital fingerprint, the agents identified a new algorithm that theoretically improves on the previous best attack by about 1,600 times for 1024-bit RSA keys. Researchers checked the method in a sandbox against an already solved RSA-1024 benchmark and said the AI-generated implementation used 3.5 times less CPU time, and less than one-thousandth of the computing power needed to factor the benchmark key itself. That run forged a signature root without recovering the private key, and the 1,600x figure comes from theoretical estimates rather than a full-scale test. Both working papers, published November 5 and 6, have not yet been peer reviewed.

Did the work also make batch attacks cheaper?

Yes, according to the same November 6 coverage. The second finding targets batch factorization of multiple RSA keys of the same bit length, which lets an attacker spread expensive work across many keys that use a standard size. CryptoBriefing's November 6 report said the new algorithm lowers the cost of that group attack. Interesting Engineering put the gain at about 30 times less average computing cost for each additional 1024-bit key versus earlier batch techniques.

The study was led by Dj Dvijotham and Abhradeep Thakurta, working with researchers from the University of California San Diego and France's Inria, including Nadia Heninger and Emmanuel Thomé. DeepMind Vice President and Chief Scientist Pushmeet Kohli announced the results. The findings do not change the advice in our what is two factor authentication guide: everyday logins still need strong modern cryptography plus a second factor.

Are today's banking and messaging systems at risk?

The researchers say no. Online banking, messaging apps, software updates, and digital currencies that rely on 2048-bit or larger keys are outside the scope of these attacks. The larger point is speed: AI agents compressed work that used to take years of manual cryptanalysis.

That same acceleration theme appeared when we covered GPT-6 Cyber.

It also shows how what is an AI agent systems can run long research tasks with human oversight.

Finding	What the November 6 reports say
Scope	Legacy RSA-1024 only; 2048-bit and larger keys unaffected
Signature padding	Little extra protection if an attacker controls about half or more of the signed block
Narrow 32-byte case	Theoretical attack about 1,600 times faster than the previous best
Sandbox check	3.5 times less CPU time on an already solved RSA-1024 benchmark
Batch factorization	About 30 times lower average cost per extra 1024-bit key