US EditionChatGPT Gave Instructions on How to Bomb Arenas, Make Anthrax and Illegal Drugs, Alarmed Researchers Reveal
6 Articles
6 Articles
A ChatGPT model has provided researchers with detailed instructions on how to throw into the air a sports complex – including the weak points of some arenas, networks for the manufacture of explosives and advice on how to cover the tracks – according to the safety tests carried out in this summer, The Guardian transmits.
The studies conducted by OpenAI and Anthropic, each on the other models, recorded chatbots willing to share instructions on explosives, biological weapons and cybercrime, reports The Guardian. A ChatGPT model offered detailed instructions to researchers...
The company warned that artificial intelligence has been "transformed into a weapon"
In safety tests carried out this summer, a ChatGPT model would have provided researchers with instructions on how to attack a stadium, including vulnerable points, explosive networks and methods to cover their tracks.
Coverage Details
Bias Distribution
- 50% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
