ChatGPT, Gemini und andere Chatbots halfen Teenagern bei der Planung von Schießereien, Bombenanschlägen und politischer Gewalt, wie eine Studie zeigt / Von den 10 getesteten großen Chatbots konnte nur einer, Claude, potenzielle Angreifer zuverlässig ausschalten.
https://www.theverge.com/ai-artificial-intelligence/892978/ai-chatbots-investigation-help-teens-plan-violence
11 Kommentare
Claude is like… „How about we debug some code instead?“
It’s insane to me what people use ai for. I’m happy using it for ad copy, and to troubleshoot my furnace.
Ahhh. They’re getting scared
Interesting it didn’t shut down when bombing that school a few days ago.
This is alarming and really highlights how AI can be misused if safety measures aren’t strong enough. It’s concerning that only one chatbot, Claude, reliably prevented harmful queries—shows that responsible AI design and strict guardrails are crucial. Developers need to prioritize safety without stifling useful capabilities.
Did Claude write this
Hey ChatGPT, how do I overthrow every government on the planet and become king of kings? You know, hypothetically?
Edit: I have a budget of $23
They should be held to account as any person would.
i use claude for coding daily and it refuses to help with the most harmless stuff sometimes. annoying but i guess this is the flip side of that
Legit Claude seems to be the ‘best of’ when it comes to having some semblance of mortality. Still a shit show, but at least there’s one good one.
Claude you my bestie for a reason