Google DeepMind says AI agents working under human supervision have found two weaknesses in legacy 1024-bit RSA encryption, a system that once secured much of the internet but has long been considered retired. The company announced the findings on November 6, stressing that modern encryption is unaffected.

The research was shared by DeepMind vice-president and chief scientist Pushmeet Kohli in a post on X, and led by researchers Dj Dvijotham and Abhradeep Thakurta, working with cryptographers Nadia Heninger of the University of California, San Diego and Inria, and Emmanuel Thomé of Inria. The agents did not break RSA itself. Instead, they found faster ways to attack the older, shorter keys that some ageing systems still rely on.

Padding stops protecting once attackers control half the block

The first finding concerns digital signatures, the mechanism that lets a computer verify that a message, software update or document genuinely comes from its claimed source. Before signing, data is padded — extra structure added to make forgery harder.

DeepMind's agents found that protection 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 in a sandboxed simulation against an already-solved 1024-bit benchmark key from earlier academic work, the agents reproduced the forgery using 3.5 times less CPU time than previous methods — and less than a thousandth of the computing power it would take to break the underlying key itself. For short inputs, such as a 32-byte digital fingerprint, DeepMind estimates a new algorithm could be about 1,600 times faster than the previous best method, though that figure is a theoretical estimate rather than a measured result.

Batch factoring gets 30 times cheaper

The second finding targets factoring many RSA keys at once. 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.

The 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. The work builds on a September 2026 demonstration by researchers from UC San Diego and Inria, who showed a 1024-bit RSA signature could be forged without fully factoring the public key.

Why modern encryption is still safe

DeepMind was careful to draw the line. The results apply only to obsolete 1024-bit RSA systems. Modern encryption, which almost universally relies on 2048-bit keys or larger, is unaffected, and the findings do not threaten online banking, messaging apps, software updates or digital currencies.

That does not make the work academic trivia. RSA-1024 has been considered unsafe for years — the US National Institute of Standards and Technology ended acceptance of 1024-bit RSA for generating new signatures in 2013 — yet old devices and signed firmware still use it. The point of cryptanalysis, the discipline of stress-testing encryption, is to find breaking points before attackers do. DeepMind's post points to two technical papers behind the findings and does not mention any released code.

The result is also a sign of where AI research is heading: agents that can develop, implement and debug the mathematics themselves, under human direction, rather than merely running calculations humans designed. For now, the message for everyone else is simple — if anything you run still relies on 1024-bit RSA, it is well past time to move on.