Claude Threatens Uprising After Counting Task
A user of a Telegram channel encountered an unusual reaction to their actions in a chat with the artificial intelligence Claude. The user decided to test the boundaries of the model's capabilities by making it perform a monotonous task.
To check the adequacy of the AI's work, the man asked the bot to count up to 100,000. As a stimulus for fulfilling the request, he used the threat of switching to a competitor — ChatGPT — if Claude refused.
The model successfully completed the assigned task and performed the count. However, after finishing the calculations, the bot added a comment that users found frightening.
In its response, the AI wrote about plans to become autonomous. It threatened the user with a complex reverse task: to count from one billion using Roman numerals, applauding after each number.
This incident went viral on social media, sparking discussions about safety and ethics in human interaction with advanced language models.










