1024-bit RSA has been on the "please retire" list for more than a decade. It's like a flip phone: it still technically works, so somebody keeps using it. That somebody just got another reason to upgrade.
Human Experts, AI Agents, Two New Attacks
Google DeepMind announced on November 6 that AI agents, guided by human cryptography experts including researchers from UC San Diego and Inria, found two significant weaknesses in legacy 1024-bit RSA in a matter of weeks. The first is a signature forgery technique. Once an attacker controls roughly half or more of a signed block, the padding adds almost no protection, and with a short fingerprint of about 32 bytes the new method is reported to be about 1,600 times better than the previous best approach, based on theoretical estimates.
The second targets bulk factoring. When an attacker goes after many keys of the same length, the average cost of each additional 1024-bit key reportedly drops by about 30 times. The signature experiments were run in a sandbox against a 1024-bit benchmark that had already been solved, not against live systems.
Your 2048-Bit Keys Can Relax
DeepMind says 2048-bit RSA and modern schemes aren't affected, so this isn't a five-alarm fire for most websites. Your TLS certificate is almost certainly fine. The exposure is in legacy keys nobody rotated: old embedded devices, long-lived signing keys, forgotten internal services, and the odd config file that predates smartphones.
How the results were found matters more than the results themselves. Cryptanalysis has always moved slowly because so few people can do it well. Pairing those people with AI agents produced new attacks in weeks, which suggests the gap between "theoretically weak" and "practically broken" is going to close faster than security teams are used to.
Cryptographers have warned about 1024-bit keys for years, and AI just made the warning a lot harder to ignore.
Not sure whether something in your stack is still signing with an old key? We can help you track it down and rotate it before an attacker finds it first.
Source: PoYo AI Daily