OpenAI’s AI Built 948 Linux Exploits, and Crypto Experts Say Decentralization Is a Shield

OpenAI says one of its AI systems dug up 948 security holes in the core of Linux and produced a working exploit for every one, at roughly $1,279 per flaw. Because most blockchain infrastructure runs on Linux, the claim matters well beyond the tech world.

How the AI Vulnerability Research Works

According to the company, the bugs let someone with a basic user account take full control of a machine, and the exploits worked on a standard desktop with normal protections switched on. The system scans code for risky areas, tests suspected flaws in a sandbox, checks whether they can be exploited and drafts a patch. Human researchers review everything before it reaches Linux maintainers. OpenAI values the bugs at $18.9 million to $66.3 million based on hacking contest payouts like Pwn2Own, against about $1.2 million in running costs. The results are self reported, and the model has not been released.

Why Crypto May Cope Better Than Most

About two thirds of Ethereum nodes run Linux, per ethernodes.org. Polygon Labs CTO Mudit Gupta says decentralization brings varied setups and redundancy that make crypto tougher against such bugs. Researcher Caducus argues constant attacks, driven by bounties, have already hardened the industry.