Google DeepMind AI Discovers New Weaknesses in Legacy 1024-bit RSA Encryption

The Google DeepMind Science team used AI agents to uncover flaws in 1024-bit RSA schemes. The team identified two weaknesses, one making signature forgery approximately 1,600 times more efficient and another cutting the average cost of factoring keys in large batches by about 30 times. The findings do not affect the security of modern systems such as RSA 2048.