A cyber-security start-up used artificial intelligence to find a critical flaw in the Java programming language, the BBC reported on Wednesday, November 18, 2026. Oracle, which owns Java, released a critical patch update on Tuesday that it said includes a fix for the flaw.

What was found

The firm, Calif, found the bug in Java's memory functions and named it Java4Shell, according to the BBC. The flaw let an attacker remotely run unwanted commands in Java's memory system. The BBC said Calif used OpenAI's recently released model GPT-6 Cyber, which is not yet available to the public.

The BBC said the bug had existed for more than a decade. It said it is not yet clear whether any platforms or companies have been affected.

Why it matters

Java was first released in 1995. Nearly all of the world's top companies use it, as do many banks, hospitals and governments, the BBC reported. Computer scientist Professor Toby Walsh told the BBC that the flaw was unlikely and serious, because Java is meant to make systems safer by managing memory. Cyber-security expert Dr Pedram Hayati said the flaw could put critical financial systems and customer data at risk.

Calif earlier found a flaw in the PHP programming language. Professor Alan Woodward of the Surrey Centre for Cyber Security said that flaw affected some 30 million websites, according to the BBC. Walsh said AI now exposes more flaws much more quickly.

In a separate article, the BBC quoted Xander Van Raemdonck, a security engineer at Aikido Security. He said many companies do not have a full list of where Java runs in their software. Sheila Pancholi of RSM UK said organisations need to act quickly when flaws appear and test that their response worked.

The BBC noted that AI firms such as Anthropic and OpenAI now give their most advanced models first to selected companies and institutions. This lets security professionals find flaws before criminals do, the BBC said.