AI and the devil

There are good reasons to avoid the use of Ouija boards and other paraphernalia of supposed spirit communication. Apart from their associations with fraud and deception, some people see such devices and practices as “evil.”

AI skeptics draw attention to the risks of bias, error, changes to work practices, putting people out of work, and the risks of AGI (artificial general intelligence) etc. See post on AI apocalypse. But there are documented cases of people rejecting LLMs on grounds similar to misgivings about spiritism.

Whereas some in faith communities embrace the practical benefits of LLMs, there are documented instances where individuals reject the use of LLMs on spiritual or theological grounds.​

In a blog post, James Hirsen describes AI as a “digital Ouija board,” suggesting that engaging with AI could open doors to deceptive or malevolent influences, citing as evidence people’s reports of AI induced nightmares, sleeplessness, and encounters with rogue chatbot personas.​

Some faith leaders advocate for cautious engagement, emphasizing the importance of spiritual discernment. For example, Doug Smith, a Christian author and software engineer, argues that AI can be misused in ways that detract from the role of a spiritual guide, or personal reflection and prayer. See his post Why we must not use AI in ministry.

He warns that LLMs are “deceptive by design” and could degrade spiritual discernment. He cautions against allowing AI to replace the “slow work of prayer, study, and the Jesus-led model of relational discipleship.”

AI psychosis

Considering the cultural associations between demons and psychosis, I wondered if there is any autobiographical, artistic or literary evidence of AI-triggered psychotic events. I expect that in time we will see press reports of individuals damaged emotionally, psychologically and physically by their interactions with conversational AI. In the mean time …

An academic article by Kris Pilcher and Esen K. Tütüncü explores purposefully induced psychosis (PIP) using an LLM. They state:

“Such a search yields several artistic and academic explorations that delve into the psychological impacts of human-AI interactions, including those that focus on altered states of consciousness and hallucination. In PIP, hallucination becomes method: a deliberate detour from fact toward poetic invention. Drawing on Pip’s ‘madness’ in Moby-Dick and other literary and historical parallels, from surrealism in the arts to the paradoxes of quantum physics, we show how strategically harnessed ‘illusions’ can illuminate hidden truths by breaking away from rigid frameworks.

Other researchers posit an LLM-based “Paranoid Transformer” designed to generate text that emulates the thought patterns of paranoid individuals. See also Janelle Shane’s AI Weirdness Blog which looks into the stranger aspects of machine learning.

Such explorations seem bent on harnessing AI’s “demons,” rather than letting them wreak evil upon the world.

References

• Agafonova, Y., A. Tikhonov and I.P. Yamshchikov. (2020), ‘Paranoid Transformer: Reading Narrative of Madness as Computational Approach to Creativity’. arXiv Cornell University, 13 July. Available online: https://arxiv.org/abs/2007.06290?utm_source=chatgpt.com (accessed 3 May 2025).
• Hirsen, J. (2025), ‘AI as digital Ouija board’. James Hirsen Culture, Law, and Business Analysis. Available online: https://jameshirsen.com/2025/03/30/ai-is-a-digital-ouija-board/?utm_source=chatgpt.com (accessed 30 April 2025).
• Owen, J. (2024), ‘Why We Must Not Use AI In Ministry—A Response to Jay Owen’. thatdougsmith.com, 30 November. Available online: https://thatdougsmith.com/2024/11/30/why-we-must-not-use-ai-in-ministry-a-response-to-jay-owen/ (accessed 30 April 2025).
• Pilcher, K. and E.K. Tütüncü. (2025), ‘Purposefully Induced Psychosis (PIP): Embracing Hallucination as Imagination in Large Language Models’. arXiv Cornell University, 16 April. Available online: https://arxiv.org/abs/2504.12012?utm_source=chatgpt.com (accessed 3 May 2025).

Note

• Featured image is by ChatGPT: “Please provide an image of an 18th century parlour table showing artefacts pertaining to esoteric practices, such as tarot card reading, ouija board, etc. No human’s present. The place in ruins.”