OpenAI says its AI found 948 Linux kernel flaws and wrote working exploits, and crypto runs on Linux

Published Thu, 05 Nov 2026 08:26:26 UTC · Updated Thu, 05 Nov 2026 20:15:00 UTC · Edited by Bizzle · AI-assisted

⚪NeutralThe research raises security questions for Linux-based crypto infrastructure but reports no attack on any network, so its impact on coins is unclear.News sentiment, not financial advice. It reflects the likely impact of this news, not a price prediction.

Summary

OpenAI said its autonomous security agents found 948 distinct flaws in the Linux kernel, wrote working exploits for them and drafted fixes for human review, for about $1.2 million in API credits, or roughly $1,279 per flaw, CoinDesk and Decrypt reported.

The flaws are local privilege-escalation bugs, meaning an attacker who already has limited access to a Linux machine could gain root and take full control; CoinDesk said OpenAI reported they worked on a standard desktop with normal protections on.

OpenAI compared the cost with historical $20,000 to $70,000 hacking-contest payouts for such exploits, putting the findings' hypothetical value at $18.9 million to $66.3 million, and framed the work as defensive, finding bugs before attackers do.

OpenAI didn't mention crypto, but CoinDesk noted much blockchain infrastructure runs on Linux, including about two-thirds of Ethereum nodes per ethernodes.org; security voices it quoted said decentralization leaves crypto better placed than most to cope.

Key facts

Decrypt said the agents scanned about 2.8 million lines of kernel code, roughly 7% of the project, reproducing 1,724 vulnerabilities from 2,308 candidate findings. (Decrypt)
An earlier July campaign found 166 unique privilege-escalation flaws for $513,690, while a newer, unreleased cyber model found more than five times as many in October, Decrypt reported. (Decrypt)
Polygon Labs CTO Mudit Gupta: "Decentralization brings diversity in infrastructure and redundancy which make crypto much more resilient to bugs like this than traditional software." (CoinDesk)
CoinDesk noted the results are OpenAI's own and the model behind them hasn't been released; Decrypt said OpenAI is limiting access to its stronger cyber models through verified security programs. (CoinDesk)