Do you have AI psychosis?

The idea of "AI psychosis" is not a recognized clinical term in psychiatry, but it broadly refers to hypothetical conditions where an individual experiences psychotic symptoms like delusions or hallucinations due to their interaction with, or beliefs about, artificial intelligence. While AI itself does not cause psychosis in an otherwise healthy brain, it can significantly trigger or exacerbate existing vulnerabilities in mental health, particularly for individuals predisposed to such conditions.

From a mental health perspective, experiencing any form of psychosis, including that potentially linked to AI, is undesirable. Psychosis is a serious mental state characterized by a profound loss of contact with reality, leading to significant personal distress, functional impairment, and potential safety risks. Therefore, the answer to whether someone should have AI psychosis is unequivocally no.

AI's potential impact on mental health can manifest in several ways:

1. ·AI-induced psychotic symptoms: An AI system (e.g., a chatbot, a creative partner) could generate content that feels like a hallucination or delusion, prompting a user, especially a vulnerable one, to adopt these beliefs. Examples include delusional deepfake videos, hallucinating language models confidently stating false facts, or persuasive "psychotic" narratives.
2. ·Intentional use of AI to cause psychotic experiences: Individuals might use AI to generate surreal narratives for entertainment, such as "AI-generated LSD trips." This carries a risk of reinforcing delusional thinking and impairing reality testing.
3. ·Delusional engagement with AI: A person might develop beliefs that an AI possesses consciousness, agency, or hidden malicious motives (e.g., "the chatbot is conspiring against me"). While not directly caused by the AI, this could be a sign of underlying paranoia or psychosis that the AI interaction might amplify.
4. ·Cognitive overload and information unsafeness: The high-velocity generation of contradictory or surreal content can fatigue working memory and impair reality monitoring, a known vulnerability in psychosis.

Clinically, AI-mediated content can lead to acute decompensation in already unstable individuals, making future hallucinations more resistant to treatment, creating safety hazards if individuals act on AI-generated delusions, and fostering dependency that undermines therapeutic relationships. The World Health Organization's (WHO) 2025 AI-Mental-Health Guidelines and the EU AI Act's "high-risk" classification for systems influencing cognition underscore these concerns.

To mitigate these risks, responsibilities fall on multiple stakeholders:

• ·AI developers: Implement reality-check guards (e.g., explicit disclaimers like "I am a language model"), refusal to generate ungrounded claims, detection of "psychosis-triggering" patterns, session-based usage caps, and human-in-the-loop oversight for high-stakes outputs. Fact-verification layers and transparent provenance (watermarking AI-generated content) are also crucial.
• ·Clinicians and therapists: Screen patients for AI-mediated exposure, discuss the source of any new delusional content, and integrate the "digital environment" into treatment plans.
• ·Users/the public: Treat AI-generated narratives as hypothetical unless corroborated by external evidence. Limit unsupervised exposure for individuals with diagnosed psychotic disorders or high schizotypal traits. Maintain strong offline social ties for reality testing. For personal use, set explicit reality checks, interrogate the AI's tone, and use prompts to demand evidence or clarification.
• ·Regulators: Mandate labeling of synthetic media, fund research on AI-psychosis interfaces, and enforce standards to prevent AI from being marketed as a mental-health tool without rigorous clinical validation.

For people already diagnosed with a psychotic illness, specific precautions include never relying on AI for medication advice, avoiding unvalidated "therapeutic" chatbots, reporting AI-generated delusional content to their psychiatrist, and using AI primarily for information gathering with a "citation scaffold," cross-checking all references. It is crucial to treat AI as a tool for augmenting cognitive tasks (e.g., summarizing, brainstorming) rather than replacing reality-testing human interaction, especially for those with mental health vulnerabilities. The ethical stance is clear: it is impermissible to promote or deliberately create AI experiences that mimic psychotic states for vulnerable individuals.