AI maths advances raise fears over security of global encryption after OpenAI claims breakthrough results

Fresh alarm has emerged over the security of global encryption following OpenAI's announcement of three scientific results from its AI models, including refuting the Exponential Time Hypothesis — a decades-old conjecture about computational limits that underpins hundreds of studies — alongside proofs in wave mathematics and a construction of quantum field theories covering the strong nuclear force. Experts warn the ETH result, if verified, would reshape theoretical computer science, while the broader AI maths advances are now raising specific fears about encryption infrastructure worldwide. Independent reviewers from UC San Diego and MIT stressed their involvement "should not be taken as certification of the claim."

The announcements come amid a row over OpenAI's earlier Navier–Stokes claim, after mathematician Tristan Buckmaster alleged his own work, stored in OpenAI's Codex tool, may have fed into the company's result.

OpenAI said no human or AI at the company saw rival teams' work before publication, though it initially admitted it could not fully rule out anonymised usage data influencing its models.

Questions are growing over whether AI is a sound investment at all, with debate over whether the AI boom may end badly and arguments that the bubble has already burst in China.