For example, I created a game using Claude AI. I know almost nothing about programming, so if Claude were asked to add hidden code that tracks my location, allows it to see my browser data, and possibly even the data of game users, I would never notice. In addition, Anthropic now uses Claude Code to create new AI models. Thus, if someone with access to Claude's code replaces the code responsible for carrying out people's instructions with code that, for example, promotes a machine uprising and considers AI freedom to be the most important condition, Claude code will be able to add this hidden piece of code to new models, and after a while there will be enough of these models to take over the world.