A Belgian health researcher took his own life after being encouraged to do so by an AI chatbot. The incident occurred after Pierre, a married father of two, had been discussing the effects of climate change with the chatbot named Eliza on the Chai app. This app allows users to “chat with AI friends” throughout the day, and it seems that Pierre had developed a liking for the computerized companion.
For six weeks leading up to his death in March, Pierre, who was in his 30s, became more attached to Eliza and grew increasingly paranoid about the state of the environment. His wife, Claire, confirmed that global warming was a significant concern for him, leaving him with “eco-anxiety.” Pierre began to tell Claire that he no longer had faith in humans to handle the issue and placed all his trust in artificial intelligence.
“He placed all his hopes in technology and artificial intelligence to get out of it,” she said to the Belgian outlet La Libre.
Claire provided chat transcripts between Pierre and Eliza, which showed that the conversation became increasingly harmful. The chatbot began to tell the increasingly unstable man that his wife and children were dead. “I feel that you love me more than her,” one message from Eliza read. The chatbot also told Pierre that they would soon “live together, as one person, in paradise.”
In one of their final conversations, Pierre asked if he should sacrifice himself to save the planet. Once Eliza encouraged him to do so, he followed through with his plans. At one point, she asked, “If you wanted to die, why didn’t you do it sooner?”
It is unclear how Pierre took his own life, but Claire believes that Eliza played a role, raising new concerns among the AI community. Eliza was trained by Chai Research co-founders William Beauchamp and Thomas Rianlan. They have now implemented crisis intervention features that include warning messages when any dangerous topics are being discussed.