The first shi virus is already here: how hackers attack computers with neurot
Although the virus is still in the concept stage and is not used in real attacks, its architecture is a dangerous precedent for the development of harmful programs, Cyber Security News writes. Unlike ordinary extortionists, Promptlock does not have a pre -compiled viral code. Instead, he sends a locally-working neural network, such as Openai GPT-OSS-20B, ordering her to generate scripts for attack.
For example, he may ask AI to write a code for collecting information about the system (username, OS type, etc. ), as well as scan files on the disk and find personal data in them. Using Lua light language makes these scripts versatile - they can work on Windows, Linux and MacOS. Technology dramatically changes the rules of the game.
Since the harmful code is created on the fly directly on the victim's device, it becomes much more difficult to detect antivirus - because there is nothing suspicious from the beginning in the file. After collecting data, the virus uses a 128-bit SPECK cipher to encrypt files. Researchers say that the virus is still in development: for example, the function of destroying data in it is prescribed, but not yet implemented.
Interesting detail: one of the instructions found a bitcoin address attributed to the creator of Bitcoin Satoshi Nakamoto, which is probably just a joke or a distracting maneuver. Earlier, Focus wrote that a dangerous virus was found in Bi browsers: what vulnerability steals data. New browsers with AI, such as Perplexity Comte, have a dangerous vulnerability that enables hackers to steal personal data, including password and bank cards information.