Source time Edition date
Google DeepMind Says AI Agents Found Two Flaws in 1,024-Bit RSA | TokenPost
Google DeepMind称AI智能体在1,024位RSA中发现两个缺陷 | TokenPost
Google DeepMind reports that human-guided AI agents surfaced weaknesses in 1,024-bit RSA, but the finding is bounded: 2,048-bit systems used today remain unaffected.
Judgment
Security teams deciding how to allocate cryptography review capacity should treat this as a reason to run human-guided AI stress tests against legacy 1,024-bit RSA, while keeping 2,048-bit systems as the production baseline; the tradeoff is spending scarce review time on older systems that the source says are not the modern standard.
Basis
Google DeepMind said human-guided AI agents found two weaknesses in traditional 1,024-bit RSA cryptosystems, and that those two vulnerabilities do not affect modern 2,048-bit systems. The mechanism described is AI-assisted stress testing: human-guided models accelerate testing of cryptographic systems, which the company frames as a way to surface weaknesses faster. DeepMind also said cryptographic standards need continued stress testing as AI advances mathematics and computing.
Unclear The source does not say which two weaknesses were found, how long the testing took, or whether any 1,024-bit system in use was affected.
Next Check whether DeepMind or an independent cryptography group publishes a technical write-up naming the two weaknesses and confirming the 2,048-bit scope.