Oracle has shipped a patch for a critical Java vulnerability that security researchers found with the help of artificial intelligence. The flaw is tracked as CVE-2026-114938 and has picked up the nickname Java4Shell.
The bug lets an attacker “write once, RCE everywhere,” according to the researchers who discovered it. That phrase means one exploit can run malicious code on every unpatched Java system it reaches, without rewriting it for each target.
The flaw can also “reveal sensitive memory,” which may expose credentials, session tokens, or other secrets held inside a running application. Oracle has released a fix, and the window for applying it is short.
Why Java4Shell reaches far beyond one server
Java runs quietly inside an enormous amount of business software. The affected products named by the researchers include Keycloak, an identity and single sign-on system, and Atlassian Jira, a project tool used by most large engineering teams.
An exploit that works across both is not a niche problem. Identity servers hold the keys to everything else, so a memory leak there can unlock far more than a single application.
AI found the bug before humans did
The same research team found a critical PHP bug last month, again with AI assistance. MyListingo covered how a PHP flaw knocked Dutch and EU government sites offline while operators scrambled to patch.
Model-driven code review is turning vulnerability hunting into a much faster process. Machines can read huge codebases in hours and flag suspicious memory handling that human auditors might miss for years.
That speed cuts both ways. Defenders get an early start on patches, but attackers work from the same models and can scan for unpatched servers within minutes of a disclosure.
What administrators should do now
Update Java runtime installations as soon as possible. Anyone running Keycloak, Jira, or another Java-based service should treat this as urgent rather than routine maintenance.
Check whether your hosting provider already applied the update. Managed platforms often patch quietly, but self-hosted deployments are usually left to the operator.
Reviewing logs for unusual outbound connections is also sensible. Remote code execution flaws rarely announce themselves, and quiet memory probing can look like ordinary traffic.
The bigger pattern
This is the second AI-assisted critical disclosure in a month, and it will not be the last. Security tooling is moving in the same direction, including the cybersecurity push inside Google’s Gemini line.
The race is now between automated discovery and automated patching. Teams that still treat updates as a monthly chore will keep losing ground to teams that can ship fixes in hours.
Read the original report at The Verge, which links to New York Times coverage of the disclosure.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.










