Google DeepMind AI Finds Legacy RSA Weaknesses 1,600 Times Faster
In a landmark cryptography announcement on November 6, 2026, Google DeepMind revealed that its human-guided AI agents had uncovered two significant weaknesses in legacy 1024-bit RSA encryption. The work took just weeks—something that could have taken human cryptanalysts years.
Led by DeepMind scientists Dj Dvijotham and Abhradeep Thakurta, in collaboration with cryptanalysts Nadia Heninger and Emmanuel Thomé, the research identified two distinct vulnerabilities:
- Signature forgery: Roughly 1,600 times more efficient against legacy RSA systems.
- Batch factoring: Approximately 30 times lower marginal cost for factoring additional 1024-bit RSA keys.
Critically, DeepMind VP of Science Pushmeet Kohli clarified that modern RSA-2048 and larger key sizes remain secure from these specific exploits. The company framed the research not as a cause for alarm, but as a demonstration of how AI can help stress-test and strengthen cryptographic standards—and as a call for organizations still using legacy 1024-bit infrastructure to upgrade urgently.
Source: Crypto Briefing – Google DeepMind’s AI Agents Expose Two Weaknesses in Legacy 1024-Bit RSA