FreshSynapse

AI trends & daily updates, in plain English

AI Research

DeepMind's AI Agents Find Two Cracks in Legacy RSA-1024 Encryption

By FreshSynapse Editorial · · 3 min read

A person typing at a keyboard in a dimly lit room in front of two monitors showing green text

Photo: Hippopx (public domain). Illustrative image.

What happened. Google DeepMind announced on Friday 6 November 2026 that AI agents, working under human supervision, uncovered two weaknesses in legacy 1024-bit RSA cryptosystems. The findings were shared by Pushmeet Kohli, DeepMind’s vice president of science, and written up in two technical papers. The work was led by DeepMind researchers Dj Dvijotham and Abhradeep Thakurta, working with cryptography specialists Nadia Heninger of UC San Diego and Inria, and Emmanuel Thomé of Inria.

The first crack: signatures you can fake. RSA digital signatures work like a wax seal, proving a message genuinely comes from its claimed source and has not been tampered with. Before signing, many standards add “padding” — extra filler data that rounds a message out to a fixed-size block. DeepMind’s agents found a sharper answer to how much protection padding actually gives: it depends on how much of the signed block an attacker controls. Once an attacker controls roughly half the block or more, padding stops adding meaningful protection, and forging a signature becomes nearly as easy as if there were no padding at all.

Working entirely in a sandboxed simulation, against an already-solved 1024-bit benchmark key from an earlier academic study, the agents’ implementation reproduced the forgery using 3.5 times less CPU time — and at under a thousandth of the computing power it would take to break the underlying key itself. For the harder cases, where an attacker controls only a small slice of the block, such as a 32-byte digital fingerprint, the agents produced a new algorithm that DeepMind’s theoretical estimates put at roughly 1,600 times faster than the best previously known method for 1024-bit keys. That 1,600x figure is a theoretical estimate, not a measured result; the measured number is the 3.5x CPU saving.

The second crack: factoring in bulk. RSA’s security rests on a simple asymmetry: multiplying two large prime numbers is easy, but splitting the product back into its original primes is extraordinarily hard. Because RSA keys typically share a standard size — 1024 or 2048 bits — an attacker targeting many keys can, in theory, spread one large upfront computation across all of them, lowering the average cost per key. DeepMind’s agents discovered a new algorithm with better asymptotic complexity than any previously known batch method, cutting the average cost of factoring each additional 1024-bit key by about 30 times.

Why it matters — and why it does not threaten your bank account. The results apply only to obsolete 1024-bit RSA. Modern encryption almost universally uses 2048-bit keys or larger, and DeepMind is explicit that those are unaffected by these specific findings. Online banking, messaging apps, software updates and digital currencies are not threatened, and no live keys were broken: the experiments ran in simulation, on a standard the US standards body NIST stopped accepting for generating new signatures back in 2013. RSA-1024 is a safe proving ground precisely because it is retired — complex enough to be a meaningful test, safe enough to attack without touching anything live.

The significance is different. Nobody should still be using 1024-bit RSA, but old devices and signed firmware still do, and the work shows that the old lock is weaker than the textbooks assumed. It is also a demonstration of a new research method: AI agents, guided by human cryptographers, developing, implementing and debugging the computational procedures behind a genuine advance in cryptanalysis.

Sources

  • Google DeepMind
  • RSA
  • encryption
  • cryptography
  • AI agents
  • cybersecurity

Related articles