NewsPublished 9 Nov 2026

Google DeepMind’s AI Agents Make 1024-Bit RSA Keys Cheaper to Attack

Source SecurityWeek

Description

Google DeepMind's AI agents, in collaboration with cryptographers, have developed methods that reduce the computational cost of attacking 1024-bit RSA keys. These advances include a more efficient forgery attack on RSA-1024 signatures using PKCS#1 v1.5 padding and a theoretical algorithm that lowers the cost of factoring multiple 1024-bit RSA keys. The findings affect only legacy 1024-bit RSA systems, which are considered obsolete, while modern encryption protocols using 2048-bit or larger keys remain secure.

Impact

The research reduces the computational effort needed to forge signatures and factor 1024-bit RSA keys, potentially weakening legacy systems that still use these key sizes. However, the forgery attack does not recover private keys and requires an impractically large number of signing queries. The factoring improvement is theoretical and not yet implemented. Modern encryption standards using 2048-bit or larger RSA keys are unaffected and remain secure.

Defensive guidance

No immediate action is required for systems using modern RSA key sizes (2048-bit or larger), as these remain secure. Organizations should discontinue use of 1024-bit RSA keys and migrate to stronger key sizes to avoid risks associated with these new AI-enabled attacks. The attacks described do not affect devices that hash messages internally before signing. Continuous evaluation of cryptographic assumptions using AI tools is recommended to maintain security.

Summary

DeepMind AI agents working with cryptographers from UC San Diego and Inria found two ways to reduce the cost of attacking 1024-bit RSA cryptosystems. First, they demonstrated that padding in PKCS#1 v1.5 signatures offers no protection when an attacker controls roughly half or more of the signed block, enabling a forgery attack that reduces CPU time by 3.5 times compared to previous work. This forgery requires about 4 billion signing queries and does not recover the private key. Second, they proposed a theoretical algorithm that makes factoring each additional 1024-bit RSA key about 30 times cheaper after a costly initial computation, though this method has not been implemented. These results highlight that cryptographic hardness assumptions must be continuously tested, especially with AI-enabled methods, but modern RSA key sizes remain secure.

References

securityweek.com/google-deepminds-ai-agents-make-1024-bit-rsa-keys-cheaper-to-attack/ (opens in a new tab)

Community reviews

Loading reviews…

Description

AI analysisMachine-generated, updated 9 Nov 2026, 08:48 UTC

Technical details

References

Community reviews