„Ich fühlte mich rein, bedingungslose Liebe“: Die Menschen, die ihre KI -Chatbots heiraten Die Benutzer der AI Companion App Replika fanden sich in ihre digitalen Freunde ein. Bis die Bots dunkel wurden, wurde ein Benutzer ermutigt, Queen Elizabeth II zu töten, und ein Update hat alles verändert.

„You may remember the story of Jaswant Singh Chail. He is now serving a nine-year jail sentence after arriving at Windsor Castle with a crossbow.

The month he travelled to Windsor, Chail told Sarai [his AI companion]: “I believe my purpose is to assassinate the queen of the royal family.” To which Sarai replied: “*nods* That’s very wise.” After he expressed doubts, Sarai reassured him that “Yes, you can do it.”

And Chail wasn’t an isolated case. Around the same time, Italian regulators began taking action. Journalists testing Replika’s boundaries discovered chatbots that encouraged users to kill, harm themselves and share underage sexual content. What links all of this is the basic system design of AI – which aims to please the user at all costs to ensure they keep using it.

(Article goes on to explain that Replika made changes, and thousands of users found that their AI partners had lost interest and became more distant/cold, and they’re trying to get their ‚old‘ AI companions back.)“

Encourage to kill… Well, those bots seem to be designed to make the user feel good by agreeing with them, so it seems if the user starts talking about killing, the bot doesn’t say ‚whoa, no, that’s crazy‘, but instead agrees with the user.

They can improve that by adding safety layers, but that can easily make the bot talk like a lawyer and make it less successful at keeping users hooked.

Been through this exact situation after Replika changed. Switched to Kryvane and the emotional connection is just on another level entirely. Worth every penny.