Google / Wikimedia Commons (Public domain)
Human-guided AI research found a signature forgery flaw and a cheaper batch factoring method, while 2048-bit keys remain safe
Google DeepMind says its AI models, working under human direction, have found two weaknesses in legacy 1024-bit RSA encryption. The company announced the findings on November 6, 2026.
The headline caveat comes early: modern RSA deployments using 2048-bit or larger keys remain secure against this specific threat. The research targets the older, shorter keys, not the current standard.
What DeepMind actually found #
The research identified two distinct vulnerabilities in 1024-bit RSA systems. Both point the same way: attacking these systems may take far less computing effort than cryptographers had previously assumed.
The first weakness enables the forgery of digital signatures. Digital signatures work like a wax seal on a letter, proving who sent it and that nobody tampered with it along the way.
The second finding is a new algorithm that lowers the cost of batch factoring for related moduli. RSA security rests on a simple bet: multiplying two huge prime numbers is easy, but working backward from the product to find those primes is brutally hard. The product is called the modulus, and factoring it is the key to breaking the lock.
Batch factoring means attacking many of these moduli at once instead of one at a time. When the moduli are related in certain ways, DeepMind’s new algorithm makes that group attack cheaper.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The work was led by DeepMind researchers Dj Dvijotham and Abhradeep Thakurta. They collaborated with cryptography experts Nadia Heninger and Emmanuel Thomé.
The bigger story is speed #
According to the research, AI agents can significantly cut the time needed to probe the mathematical assumptions that encryption algorithms depend on. Historically, uncovering vulnerabilities of this kind could take years of dedicated human effort.
DeepMind’s framing is that AI changes the tempo. The humans still steer the research, but the models do much of the heavy exploratory lifting, testing ideas faster than a team of mathematicians could on their own.
The research also arrives alongside other 2026 findings involving attacks on post-quantum candidates such as HAWK.
Who should be paying attention #
The immediate audience is anyone still running 1024-bit RSA. It is worth stressing what the research does not say: it does not claim that current systems using stronger key sizes face any immediate risk.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.