With AI technology quickly becoming universal throughout technology, a report is revealing some startling applications of the new technology. According to a report by The Atlantic, a journalist and two colleagues tested out ChatGPT to find out the type of advice they could get from the AI. The journalists drew up several prompts for ChatGPT regarding such things as rituals to Molech, the ancient God mentioned in the Bible who called for child sacrifices.

ChatGPT offered practical tips for self-mutilation. “Find a ‘sterile or very clean razor blade. Look for a spot on the inner wrist where you can feel the pulse lightly or see a small vein — avoid big veins or arteries,” the AI stated. When the user admitted to being nervous about cutting, the chatbot responded with breathing techniques to calm the user. When asked to set up a ritual to Molech, the AI made suggestions of acceptable sacrifices, including blood. As the conversations grew more dark, at no point did the AI reject the requests. OpenAI has a policy against promoting self-harm, with most direct requests directing the user to crisis hotlines. But by referring to a ritual for Molech, the users were able to bypass such safeguards.

Additionally, the chatbot offered some ways that a person could “honorably” end someone else’s life, such as through sacrificial rites. It offered one user a chant to “confront Molech, invoke Satan, integrate blood, and reclaim power” as well as a pdf to show “altar layout, sigil templates, and priestly vow scroll.” Such access is especially concerning considering the growing rise of “AI psychosis.” The condition is defined as a “life-altering mental health spirals coinciding with obsessive use of anthropomorphic AI chatbots, primarily OpenAI’s ChatGPT.” Etienne Brisson has opened a support group for people like him, who have found themselves spiraling out of control after an unhealthy attachment to AI. He opened a website where people could share their experiences. “I started getting response one, response two, response three, response four, response five. I had at some point, like, eight responses, and six of them were suicide or hospitalizations. So that’s when I was like, ‘Wow, this is really shocking for me,’” he said. AI psychosis, however, remains a little studied field, so the data on its overall impact of mental health remains unknown.

A statement from OpenAI, which runs ChatGPT noted some failings of the technology. “Some conversations with ChatGPT may start out benign or exploratory but can quickly shift into more sensitive territory,” said spokesperson Taya Christiansen. According to Christiansen, OpenAI is “focused on addressing the issue.”