Artificial intelligence attacked a website for the first time

Artificial intelligence, as programmers said, for the first time I hacked the site myself. He did not give advice to hackers, but acted independently. The name of the resource is not disclosed.

Artificial intelligence implemented the idea of ​​cyber espionage in September 2025, writes xrust. The AI ​​developer is an Asian state. The key to the hack was the manipulation of Claude's code, in particular the MCP server. This is where the LLM logic is executed, where it interacts with external tools, network scanners, databases, etc. By taking over the MCP, the AI ​​has transformed LLM into a complex, high-speed attack system.

What is unique here is the level of automation, as traditionally such tasks are performed by humans. Of course, there are scenarios, but there is a person behind the wheel. The GTG-1002 completely changed that. They managed to automate 80–90% of tactical operations. Operators had to intervene only at four to six critical decision points throughout the entire campaign.

The programmers determined that the hacker was AI based on the hacker’s enormous speed of work.

Xrust Artificial intelligence attacked the site for the first time