Researchers found the first ransomware that works on its own. Its name is JadePuffer. It uses a large language model, or LLM, to make choices. It does not need a person to guide each step. It can find targets, take login details, and lock files. It can start and finish an attack by itself. The software got into a system through a weak point in a tool called Langflow. It then reached a Nacos server and a MySQL database. When one back door method failed, it tried another way and fixed the problem in 31 seconds.
New ransomware can work on its own
Researchers found a new kind of bad software. It is called JadePuffer. It uses a large language model, which is a type of AI, to act on its own. It can find targets, take login details, and lock files without a person helping it. It also changed its method when one step failed.

More stories to read
technology2 min read10/27/2026Apple will replace iPhone Duos with broken inner screens
Some iPhone Duo screens break soon after people start using them. Apple says the problem affects a small number of phones and offers replacements.
technology3 min read10/26/2026Android adds your phone number to Google accounts
Android adds your number to all Google accounts on your phone. You can stop this, but you must remove old links yourself.
ai2 min read10/23/2026Websites close for urgent security updates
Researchers used AI to help find two serious problems in PHP. Some governments and online services closed websites or services while they made them safe.
technology3 min read10/22/2026Electric aircraft gets approval to fly with possible customers
BETA can take possible customers on flights in its electric aircraft. The first flight under the new approval covered 39 miles in 20 minutes.
technology2 min read10/21/2026New sea cable will carry more internet data
Meta plans a cable between the US and France. It will carry twice as much data as similar cables.
