Vidhay Reddy Says He Received a Threatening Response from Google Gemini’s AI Chatbot

REDDY I was definitely ░░░░░░░░ scared I was just ░░░░░░░░ the couch my sister ░░░░░░░░ watching TV on the ░░░░░░░░ so I was on ░░░░░░░░ laptop and I was ░░░░░░░░ some of this work ░░░░░░░░ then yeah it just ░░░░░░░░ all of a sudden ░░░░░░░░ 1 voice over You ░░░░░░░░ a burden on society ░░░░░░░░ are a drain on ░░░░░░░░ earth You are a ░░░░░░░░ on the landscape You ░░░░░░░░ a stain on the ░░░░░░░░ Please die REDDY It ░░░░░░░░ felt like it was ░░░░░░░░ personal attack Reporter 2 ░░░░░░░░ over CBS News Confirmed ░░░░░░░░ Google for their response ░░░░░░░░ their statement they said ░░░░░░░░ language models can sometimes ░░░░░░░░ with non sensical responses ░░░░░░░░ this is an example ░░░░░░░░ that This response violated ░░░░░░░░ policies and we ve ░░░░░░░░ action to prevent similar ░░░░░░░░ from occurring REDDY I ░░░░░░░░ there s the question ░░░░░░░░ liability of harm So ░░░░░░░░ they were to say ░░░░░░░░ this message was just ░░░░░░░░ non sensical it was ░░░░░░░░ a general AI hallucination ░░░░░░░░ it could be just ░░░░░░░░ dismissed Reporter 2 Some ░░░░░░░░ claimed that the chat ░░░░░░░░ have been manipulated either ░░░░░░░░ voice or code Reddy ░░░░░░░░ he did nothing to ░░░░░░░░ the output from Gemini ░░░░░░░░ did not reply when ░░░░░░░░ if Gemini could be ░░░░░░░░ to elicit this response ░░░░░░░░ way the response it ░░░░░░░░ violates Google s policy ░░░░░░░░ It s not the ░░░░░░░░ time a chatbot has ░░░░░░░░ alarm There have been ░░░░░░░░ instances of concern raising ░░░░░░░░ over AI safety.”