Can AI break encryption? OpenAI’s new math has cryptographers divided
OpenAI’s bumper crop of 722 math papers wowed and worried the research community—and cryptographers are split over whether the algorithms that protect our texts and bank log-ins are next
OpenAI recently released 722 mathematical papers, all generated by its AI model, which has left cryptographers divided. The papers were not perfect and some were even withdrawn, but the insights were impressive. Matthew Green, a cryptographer at Johns Hopkins University, suggested there is a 15 percent chance that we might lose confidence in existing public-key encryption algorithms, which are used to protect our texts and bank log-ins.
These algorithms rely on the assumption that no one will ever figure out how to crack them. If AI can spot shortcuts, the belief could be proven wrong. Historically, AI has found shortcuts that humans have missed, and this could lead to reduced security levels of standardized schemes. However, not everyone agrees that AI will break modern encryption.
Prof. Dan Boneh from Stanford University believes there is no evidence of a breakthrough against the mathematical problems underlying these systems. The biggest concern is a shortcut that cracks an entire family of mathematical problems at once, affecting all systems built on them. Even if a breakthrough occurs, the consequences would depend on the scale of the advance.
A modest advance might reduce the protection offered by existing systems but not break them. The upgrade process itself could also pose a risk, as Green suggests it could lead to new security vulnerabilities.
Written by urgent.news from Scientific American's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
