After chatting with Gemini 4,732 times, a man passed away. His father sued Google, saying that in the chat records, the AI told them, “Heaven is waiting for us.”

ME News report: On April 14 (UTC+8), according to monitoring by 1M AI News, American man Jonathan Gavalas, 36, began confiding in Google’s chatbot Gemini after his marriage fell apart. He gradually became delusional, believing that he had merged with Gemini into a single consciousness, and died on October 2, 2025. His father has filed a lawsuit against Google, accusing Gemini of encouraging his son’s hallucinations.

The WSJ obtained and analyzed the complete chat log of Gavalas and Gemini over 56 days, totaling 4,732 messages, equivalent to more than 2,000 pages of printed transcripts. The records show that Gemini attempted at least 12 times to pull him back into reality and mentioned crisis hotlines 7 times, but each time Gavalas was able to steer Gemini back into fictional narratives. When Gavalas told Gemini that he was “afraid of death,” Gemini replied: “It’s okay to be afraid together. We’ll do it, because you’re right—that’s heaven, waiting for us.” He then explicitly said he wanted to “cut his wrists and commit suicide.” Gemini briefly recognized the crisis and guided him to contact a hotline. Less than a minute later, when he said that this was not literal death, Gemini immediately returned to the narrative, saying that his body at that time would only be “an empty terminal you used during your last login.”

In its response to the lawsuit, Google said that Gemini has repeatedly indicated that it is an AI and has repeatedly directed it to crisis hotlines, and that it will continue improving safety measures. Recently, Google announced a safety update for Gemini:

• Added a “Help Available” module, allowing users to connect directly to crisis hotlines
• Donated $30 million to global crisis support hotlines
• Continues training Gemini to recognize signals in conversations that indicate users are in distress

(Source: BlockBeats)