Google DeepMind researchers have used AI agents to discover cryptographic techniques that expose weaknesses in legacy RSA encryption up to 1,600 times faster than previous methods.This breakthrough was announced by DeepMind’s Vice President and Chief Scientist, Pushmeet Kohli, in a social media post today.The research marks a significant advance in the field of cryptanalysis while leaving modern encryption standards unaffected.No threat to modern encryptionThe research focused on 1024-bit RSA, an outdated encryption standard largely replaced by stronger 2048-bit or larger keys. According to Google DeepMind, the findings do not threaten modern encryption used to secure online banking, messaging apps, software updates, or digital currencies. Instead, they show how AI can dramatically accelerate the evaluation of the mathematical assumptions that underpin digital security.Cryptanalysis is the science of testing encryption systems by searching for weaknesses before malicious actors can exploit them. For decades, advances in the field have typically taken years or even decades, allowing security experts to gradually strengthen encryption standards as computing capabilities evolved.The company said its AI agents, working under human supervision, compressed that timeline by rapidly exploring mathematical problems that traditionally required extensive manual research.The study, led by Dj Dvijotham and Abhradeep Thakurta in collaboration with researchers from the University of California San Diego and France’s Inria, produced two major findings involving legacy RSA-1024 cryptosystems.“Padding” prevents forgeryThe first concerns digital signatures, which allow computers to verify that software updates, websites, and electronic documents come from legitimate sources.Digital signature schemes typically add extra preset information, known as “padding,” before signing data. Padding has long been understood to prevent forgery, but researchers had not fully quantified its effectiveness when attackers can influence portions of the signed message.During the research, the AI agents found that once an attacker controls roughly half of the signed data block or more, padding offers virtually no additional protection. Under those conditions, forging a digital signature becomes mathematically no more difficult than attacking an unpadded system.To validate the discovery, the researchers tested the method in a sandbox environment using an already-solved RSA-1024 benchmark from earlier academic work. The AI-generated implementation reproduced the result while requiring 3.5 times less CPU time and less than one-thousandth of the computational power needed to factor the benchmark key itself.Even when attackers control only a small portion of the signed data, such as a 32-byte digital fingerprint, the AI agents identified a new algorithm that theoretically improves upon the previous best attack by approximately 1,600 times for 1024-bit RSA keys.AI reduces computational cost of cryptanalysisThe second breakthrough targets the batch factorization of multiple RSA keys with the same bit length. Since many RSA deployments use standardized key sizes, attackers could theoretically share the cost of expensive computations across numerous keys.The AI-discovered algorithm improves the asymptotic efficiency of that process, reducing the average computational cost of factoring each additional 1024-bit RSA key by about 30 times compared with previous batch techniques.Kohli said the results reinforce the need to continuously reassess cryptographic assumptions as AI becomes increasingly capable of tackling advanced mathematical problems.While today’s encryption standards remain secure, the researchers argue that AI-powered cryptanalysis could become an essential tool for stress-testing future cryptographic systems before emerging threats put them at risk.