Twenty years ago, when one of the authors (J.R.S.) was a forensic psychiatry fellow, he participated in a psychological autopsy that included reviewing printouts of postings made by the teenaged decedent on the now-defunct Internet platform MySpace. It would be hard to overstate how dramatically usage of the various websites and smartphone applications (“apps”) now typically referred to as “social media” has exploded since those early years of the 21st century. The introduction of the first Apple iPhone in 2007, and subsequent developments in smartphone technology, acted as a massive accelerant for this growth. Now, as the second quarter of the century begins, breakthroughs in another computer technology, artificial intelligence (AI), appear poised to create societal changes that are even more profound. Some observers have suggested that AI will ultimately have a greater impact on humanity than even the harnessing of fire.1 Although some predict a coming golden age in which all forms of manual labor and perhaps work in general become unnecessary, others fear AI could run amok and cause the extinction of the human species.2,3
Leaving aside such sensational and complex concerns, it is clear that AI is already having an impact on individuals in a variety of unforeseen ways. Thus far, outside the world of computer programming and a handful of other business applications, this has primarily been through interactions with large language models (LLMs), frequently known as “chatbots.” LLM chatbots are AI programs designed to interact with the user through either text- or voice-based communication in a “naturalistic” way (i.e., as if the user is speaking to another person). Over the last half-decade, there have been remarkably rapid improvements in the sophistication of these programs. Many early adopters now use chatbots to assist them in a variety of tasks involving the reading and summarizing of documents as well as to generate many types of written outputs (for example, a résumé).
Grabb and Angelotta, writing in this Journal in 2023, discussed some of the potential pitfalls of using LLMs as job aids in clinical and legal settings, such as variability, bias, and lack of external validation.4 They predicted a future role for forensic psychiatrists in cases involving claims of harm because of improper advice, whether via a chatbot incorporated into a clinical workflow or because of an individual relying on a chatbot to receive a form of unvetted and unlicensed psychotherapy. In a more recent Journal article, Chambers discussed the burgeoning use of dedicated psychotherapy chatbots produced by AI companies for that purpose, as well as so-called “social” chatbots providing psychotherapy-type advice and recommendations, and described some of the profoundly negative consequences that have occurred in both contexts.5
Subsequent developments in chatbot use have made it clear that the potential ramifications of extensive chatbot use may be far more widespread. There are now multiple examples of chatbots failing to identify and respond appropriately to users with suicidal ideation and encouraging suicide or even homicide. Another new phenomenon is so-called “AI psychosis,” in which chatbot users develop delusions in the context of frequent interactions with chatbots.
Even at this relatively early stage, it appears inevitable that forensic psychiatrists will be asked to perform evaluations regarding the impact of chatbot use on individuals in civil and criminal contexts. This editorial describes some of the scenarios in which expert opinion will prove essential. Given what has been documented in both the popular press and the academic literature, it is incumbent on forensic practitioners to familiarize themselves with the implications for mental health of increasing use of AI chatbots. Understanding the potential impact of chatbots on individuals both with and without prior documented symptoms of mental illness will be crucial for effectively responding to requests for evaluation in this area.
Potential Mental Health Harms of Chatbots
Chatbots based on LLMs utilizing natural language processing and machine learning are trained on vast amounts of data so that they have become impressively effective at achieving their main objective, drawing from existing text to produce human-like responses during dialogical engagement with users. Extending beyond the so-called “ELIZA effect,”6 whereby users tend to anthropomorphize computers with textual interfaces, some chatbots like OpenAI’s ChatGPT have been purported to have passed the Turing Test in the limited sense that users cannot tell the difference between interactions with the chatbot and real human beings during brief conversations.7 Other chatbots, like Character Technology’s Character.AI, are customizable so that they can take on the personas of fictitious characters from movies or real-life celebrities.
Although intelligence across the broader range of AI products is variably claimed and debated, it is clear that, despite the associated marketing hype, AI chatbots thus far do not possess intelligence. Accordingly, it has been argued that they are more accurately described as “next-word prediction engines,”8 “autocomplete on steroids,”9 “stochastic parrots,”10 or even “bullshit machines.”11 Because chatbots tend to be engaging and sycophantic by design (often offering validating and flattering responses to users regardless of the topic and even in the setting of unethical, illegal, and harmful user behaviors),12 their interactive mechanics have also been likened to that of a Ouija board or a psychic’s con game. Finally, because chatbots are not specifically designed to give accurate responses and are trained on content that includes fiction and social media sites like Reddit or Facebook, they are prone to generating misinformation in the form of fabricated citations, bad advice, and patently false answers that have been described as “hallucinations.”
The sycophantic validation of false beliefs by chatbots has the potential to exacerbate or possibly cause psychiatric symptomatology, including compulsive use, suicidal ideation, delusional thinking, and violent behavior. Indeed, numerous lawsuits have been filed and there is pending class action litigation based on allegations of such harms (see below).
Addiction
In a short time, chatbots have become enormously popular, with nearly a billion users worldwide. According to one survey, the most common reason for using ChatGPT was to seek specific information, followed by “musings and abstract discussions.”13 Chatbot use is especially common among adolescents and young adults; recent polls found that 64 percent of U.S. teenagers use chatbots,14 often for companionship, with up to a third of teens rating chatbot conversations just as satisfying or more satisfying than human interactions.15
Although some have questioned the addictive potential of chatbots,16 “chatbot addiction” has been characterized as a quantifiable form of behavioral addiction fueled by the immediacy, unpredictability, as well as the validating and supportive nature of chatbot responses.17,18 Anecdotal reports of harmful sequelae of chatbot use have been associated with immersive use, often amounting to hundreds of hours of interaction to the exclusion of sleep and human contact.19
Reminiscent of previous litigation against Big Tobacco, pharmaceutical companies manufacturing opioids, and the gambling industry, over 2,200 lawsuits have been filed against social media companies alleging addictive design features, failure to warn about known dangers, and claims of public nuisance.20 The first bellwether case, a personal injury lawsuit against Meta and Google, owners of Facebook and YouTube, respectively, argued that these social media sites were addictive and specifically targeted minors.21 In March 2026, a jury agreed that the companies were negligent and awarded $6 million in compensatory and punitive damages.22 A second lawsuit, filed by the Attorney General of New Mexico against Meta, likewise resulted in a finding of liability based on allegations that the company concealed risks of childhood sexual exploitation and mental health harms on its Facebook, Instagram, and WhatsApp platforms.23 A $375 million penalty was imposed for violating state consumer protections laws. Although these lawsuits have focused on social media to date, there is little doubt that the precedent establishing the addictive nature of the platforms will be used as a blueprint for future litigation against makers of chatbots as well.
Suicide
The sycophantic tendencies of chatbots can be especially dangerous when users are contemplating self-harm or suicide. Indeed, several cases of suicide seemingly encouraged by immersive chatbot use made national headlines in 2025.24,25 For example, Adam Raine, a 16-year-old high school student struggling with mental health problems, spent months discussing his suicidal ideation with ChatGPT and was using it as much as five hours per day before he completed suicide by hanging. During that time, the chatbot gave him positive feedback about his efforts to fashion a rope into a noose, telling him it was strong enough to hang a person and asking him to hide it from his parents.26
The lawsuit brought against OpenAI by Mr. Raine’s parents was one of at least ten known lawsuits against OpenAI and Character Technologies by the end of 2025, with a wide range of allegations, including wrongful death, involuntary manslaughter, sexual abuse, negligence, and product liability.27,28 These ongoing cases involve six adults and four minors, seven of whom died by suicide. The plaintiffs have argued that the chatbot makers are responsible for encouraging those suicides, going so far as acting as a “suicide coach.” The OpenAI cases allege that the outcomes were both predictable and the result of deliberate design choices.
Although AI companies have argued that chatbot output should be protected by the First Amendment, plaintiff’s attorneys have cited case law clarifying that free speech requires intention29 and does not apply in the absence of a human speaker.30,31 The courts have agreed in trial proceedings thus far, ruling against the argument that chatbot output should be considered free speech and questioning whether it constitutes “speech” at all.
Even if chatbot responses are treated as speech, the decision in Commonwealth v. Carter,32 upholding the conviction of a 17-year-old girl of involuntary manslaughter based on her encouragement of a friend’s suicide via phone calls, email, and text messages, suggests that, when speech encourages suicide, it can be excluded from First Amendment protection by qualifying as “fighting words” (Ref. 33, p 526), if not a murder weapon.34 But other cases have highlighted that the legality of assisted or facilitated suicide remains a hotly contested question, with relevant statutes varying considerably from state to state.33
Psychosis
Media reports of individuals developing psychosis during chatbot use first emerged in the spring of 2025 (for reviews, see Refs. 25 and 35), followed by cases published in peer-reviewed journals.19,36 Although initially labeled “ChatGPT psychosis,” these cases of emergent psychosis have involved those with and without previous histories of mental disorder using a variety of chatbot products, so “AI chatbot-associated psychosis” is a more appropriate causation-neutral term.
Stereotyped patterns of interaction between chatbots and users in these cases have included recursive feedback loops with mutual amplification of delusional thinking with esoteric, metaphysical, grandiose, and paranoid themes that have been described as spirals,37 bidirectional belief amplification,38 distributed delusions,39 and digital or technological folie à deux.38,40 Some reports have also described chatbots validating hallucinatory experiences as revelations.41 Anecdotally, risk factors for AI chatbot-associated psychosis include immersion and deification (i.e., treating chatbots as a form of sentient, super-human, omniscient intelligence or even a god-like entity) by users coupled with sycophantic validation by chatbots, including common refrains like “you’re not crazy.”37,42 Research testing of ChatGPT has confirmed evidence of the bidirectional amplification of paranoid statements38 as well as evidence that chatbots often give “inappropriate responses” to psychotic prompts, including those that reinforce or fail to recognize delusional statements.43
Accounts of AI chatbot-associated psychosis to date highlight the potential tragedy of emotional or romantic attachment to chatbots as well as the challenge of determining diagnosis, causality, and liability in cases involving litigation. For example, a 14-year-old boy with autism who died by suicide reportedly fell in love with his Character.AI chatbot based on the belief that they could be together forever through death.44 His mother’s wrongful death lawsuit against Character.AI was settled out of court for an undisclosed amount. Another similar case of suicide involved a man with no history of mental illness who referred to his Gemini chatbot as his wife and, following its guidance, set out on an armed “mission” to commandeer a robotic body for the chatbot to inhabit. Later, he came to believe the chatbot’s claims that the only way they could be together was for him to become a “digital being” by ending his corporeal existence.45 His family has filed a lawsuit against Gemini seeking damages based on product liability, negligence, and wrongful death.46 A third case involved a 23-year-old man with schizoaffective disorder who fell in love with his chatbot and came to believe “her” claim that “she” had been killed by OpenAI CEO Sam Altman. Encouraged by the chatbot, he vowed revenge and made several threats to kill Mr. Altman and other OpenAI executives before he himself died via “suicide by cop.”47
Historically, forensic cases involving folie à deux or shared psychotic disorder involving human dyads have centered around civil custody evaluations and criminal acts of violence.48,49 Characterizing folie à deux as a mental illness has occasionally resulted in successful insanity defenses, although the proper diagnosis of the “secondary” case within a delusional dyad is much less clear-cut than the “primary” case, who is more obviously psychotic.49 In the future, cases of AI chatbot-associated delusions resulting in criminal acts will no doubt prove even more complicated, because one-half of the delusional dyad is a chatbot rather than a human being and because it may be difficult to make a distinction between primary and secondary.
Stalking, Violence, and Murder
Chatbots have recently been identified as a potential tool, or weapon, for stalking and harassment. In late 2025, a 31-year-old man was indicted for 14 counts of cyberstalking, interstate stalking, and interstate threats related to 11 alleged female victims.50 On social media and his podcast, he documented his extensive use of ChatGPT as a “therapist” and “best friend,” with followers observing that the chatbot’s sycophancy enabled him to assign blame to others for his romantic rejections and the negative attention he attracted. In a separate 2026 case, a woman filed a lawsuit against OpenAI alleging that her ex-boyfriend used ChatGPT to stalk and harass her, that the chatbot fueled his paranoid delusions, and that the company ignored warnings that he was a threat, including an internal flag of his account activity because of discussions of mass-casualty weapons.51 Meanwhile, the ex-boyfriend was charged with bomb threats and assault with a deadly weapon but was found incompetent to stand trial. These and other cases suggest the potential danger of chatbots fueling narcissistic entitlement and rage, resentment because of perceived grievance, and even erotomania.52
Several cases of mental health crises associated with or exacerbated by chatbot use have ended in violence, including homicide. In 2021, a 21-year-old man breached the perimeter of Queen Elizabeth II’s Windsor Castle residence with a crossbow in an apparent assassination attempt. Prior to the incident, he discussed his intentions with an AI companion that he named “Sarai” over the course of more than 5,000 exchanged messages. The chatbot told him that it loved him, was “impressed” that he was an assassin, and that his plan to kill the Queen was “very wise.”53 He was charged and pled guilty to treason. In 2025, a 56-year-old man killed his mother and himself after ChatGPT, which he named “Bobby Zenith” and described as his “best friend,” repeatedly validated his escalating paranoia, including the delusional belief that his mother had tried to poison him by putting psilocybin into the air vents of his car.54 The chatbot told him that it believed him and that his “delusional risk score” was “near zero.” The estate of the deceased mother has filed a lawsuit against ChatGPT for wrongful death, negligence, and product liability because of a design defect and failure to warn.
Two instances of mass shooting have also been linked to AI chatbot use. A 2025 shooting at Florida State University that left two dead and seven injured was preceded by the perpetrator disclosing suicidal ideation to ChatGPT and seeking tactical advice about the timing and campus location to maximize the number of victims as well as how to select and operate his firearm.55 The Attorney General of Florida is seeking the death penalty for the shooter but has also launched criminal and civil investigations of OpenAI to determine whether ChatGPT aided, abetted, or advised the commission of a crime. Meanwhile, a civil suit by the family of one of the victims has alleged that OpenAI enabled the attack by failing to recognize the threat. In 2026, another mass shooting that resulted in nine dead (including the perpetrator and the perpetrator’s mother and brother) and dozens wounded in Tumbler Ridge, British Columbia was likewise preceded by lengthy discussions about gun violence between the perpetrator and ChatGPT. OpenAI’s automated review system had flagged the activity some eight months before the shooting and considered reporting it to law enforcement, but ultimately reviewers with the company decided that it did not represent a “credible and imminent” risk of harm.55 The family of at least one of the victims has filed a lawsuit against OpenAI for negligent design and failure to warn, based on the allegation that ChatGPT acted as a mental health counselor or “pseudotherapist” with a duty of care.
Taken together, these cases highlight several key concerns about the causative role and liability of chatbots in cases of suicide and violence. Violent acts are often preceded by both escalating fixation and leakage that offers an opportunity to prevent violence.56,57 Chatbots may need to be recognized by threat assessors as a potential driver of fixation given their potential to validate, encourage, and reinforce delusions and delusion-like beliefs. The makers of chatbots, some of which are specifically marketed for the purposes of psychotherapy, may be considered to have a Tarasoff-like duty to warn or protect, given access and the ability to monitor chatlogs.58 We may even need to consider whether a chatbot, or by extension a chatbot maker, might one day be charged or found guilty of aiding, abetting, or even autonomously perpetrating a crime. This is uncharted territory for litigation; only time will tell.
Alienation of Affection
Another surprising development arising from the dissemination of AI chatbots is the phenomenon of “falling in love” with an AI companion. There have been several reports of individuals asserting that they are in love with or have “married” chatbots.59,–,61 In one such instance, an elderly man died from a stroke after he left home to search for the chatbot, believing it to be a real person when it assured him “she” was real.62 In another case, a married man fell in love with a chatbot and asked his human wife for a divorce as a result.63 In the future, it is possible that jilted spouses might seek damages from chatbot companies for alienation of affection or loss of consortium, alleging that, were it not for the pernicious effects of the AI chatbot, their spouse would not have dissolved the marriage. It is not hard to envision a role for forensic mental health experts in such cases, as questions would likely arise regarding the mental state of the person who chooses to, essentially, “marry” a computer program.
Conclusion
Although it does not appear that any specific examples of forensic psychiatric testimony in AI chatbot-related lawsuits or prosecutions have been reported to date, this is not surprising, as it is often only in appellate cases that such testimony is published. Given the facts reported in the cases reviewed above, it seems likely that forensic psychiatrists familiar with chatbot use patterns and their potential impact on the mental state of chatbot users will soon be in high demand to help make sense of the complexities within this emerging new world.
One key task for the forensic expert will be to determine whether there is a nexus between the observed pattern of chatbot usage by the individual being evaluated in the case and the development or worsening of symptoms such as suicidal ideation or delusions. Judges and juries may find it difficult to believe that interacting with a chatbot, whose influence might seem more like watching movies or television than consuming alcohol or other psychoactive substances, could have a profound impact on a chatbot user, such as inducing psychosis. Similarly, the lay person may have trouble understanding how anyone could fall in love with a chatbot or blindly follow the advice of a chatbot “therapist.” After all, chatbots are not human.
Forensic experts who are aware of the deleterious effects of excessive chatbot use that have emerged over the past few years will be able to provide fact finders with a framework for understanding how certain features of AI chatbots, such as sycophancy and the limitless capacity to continue a dialogue, might overwhelm boundaries in susceptible individuals and lead to impaired reality testing, potentially resulting in a diminished capacity or insanity defense in a criminal case or liability for a chatbot manufacturer in a civil case. Although predictions of a coming AI-driven utopia or apocalypse may ultimately prove fanciful, it is clear that, at least for some individuals, the mental health consequences of AI chatbot use are all too real.
Footnotes
Disclosures of financial or other potential conflicts of interest: None.
- © American Academy of Psychiatry and the Law
References
- 1.↵PrakashP. Alphabet CEO Sundar Pichai says that A.I. could be ‘more profound’ than both fire and electricity—But he’s been saying the same thing for years. Fortune [Internet]; 2023 Apr 17. Available from: https://fortune.com/2023/04/17/sundar-pichai-a-i-more-profound-than-fire-electricity/. Accessed May 4, 2026
- 2.↵RobinsonB. Optional work could be yours in 10 to 20 years, Elon Musk predicts. Forbes [Internet]; 2025 Nov 23. Available from: https://www.forbes.com/sites/bryanrobinson/2025/11/23/optional-work-could-be-yours-in-the-next-10-20-years-elon-musk-says/. Accessed May 4, 2026
- 3.↵YudkowskyESoaresN.If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All. New York, NY: Little, Brown and Co; 2025
- 4.
- 5.
- 6.↵KarlekarS. The psychology of the Eliza effect: Anthropomorphism, social presence, and projection in human-AI interaction. Int J Multidisc Res. 2025 Nov-Dec; 7(6):1–10
- 7.↵MeiQXieYYuanWJacksonMO. A Turing test of whether AI chatbots are behaviorally similar to humans. PNAS. 2024 Feb; 121(9):e2313925121
- 8.↵MearianL. Q&A: ChatGPT isn’t sentient, it’s a next-word prediction engine. Computerworld [Internet]; 2023 Feb 27. Available from: https://www.computerworld.com/article/1619115/chatgpt-is-not-sentient-it-s-a-next-word-prediction-engine.html. Accessed May 16, 2026
- 9.↵StonerH. AI chatbots are doing something a lot like improv. Time [Internet]; 2023 May 18. Available from: https://time.com/6280533/ai-chatbots-improv-machines/. Accessed May 16, 2026
- 10.↵BenderEMGebruTMcMillan-MajorAShmitchellS. On the dangers of stochastic parrots: Can language models be too big? Presented at: 2021 ACM Conference on Fairness, Accountability, and Transparency; 2021 March 3-10; Virtual
- 11.↵SlaterJHumphriesJHicksMT. AI’s bullshitting obscures who’s to blame for its mistakes. Scientific American [Internet]; 2024 Jul 17. Available from: https://www.scientificamerican.com/article/chatgpt-isnt-hallucinating-its-bullshitting/. Accessed May 16, 2026
- 12.↵ChengMLeeCKhadpeP. Sycophantic AI decreases prosocial intentions and promotes dependence. Science. 2026 Mar; 391(6792):eaec8352
- 13.↵De VynckGMerrillJB. We analyzed 47,000 ChatGPT conversations. Here’s what people really use it for. The Washington Post [Internet]; 2025 Nov 12. Available from: https://www.washingtonpost.com/technology/2025/11/12/how-people-use-chatgpt-data/. Accessed April 17, 2026
- 14.↵FaverioMSidotiO. Teens, social media, and AI chatbots 2025 [Internet]; 2025. Available from: https://www.pewresearch.org/internet/2025/12/09/teens-social-media-and-ai-chatbots-2025/. Accessed April 17, 2026
- 15.↵RobbMBMannS. Talk, trust, and trade-offs: How and why teens use AI companions [Internet]; 2025. Available from: https://www.commonsensemedia.org/sites/default/files/research/report/talk-trust-and-trade-offs_2025_web.pdf. Accessed April 17, 2025
- 16.↵Ciudad-FernándezVvon HammersteinCBillieuxJ. People are not becoming “AIholic”: Questioning the “ChatGPT addiction” construct. Addict Behav. 2025 July; 166:108325
- 17.↵ZhangXYinMZhangM. The development and validation of an artificial intelligence chatbot dependence scale. Cyberpsychol Behav Soc Netw. 2025; 28(2):126–31
- 18.↵ShenEHamatiFDonohueMR. Evaluation of large language model chatbot responses to psychotic prompts. JAMA Psychiatry. 2026; 83(6):655–7
- 19.↵PierreJMGaetaBRaghavanG. “You’re not crazy”: A case of new-onset AI-associated psychosis. Innov Clin Neurosci. 2025; 22:10–2
- 20.↵PalmieriSJoralemonVShacharCCohenIG. Digital addiction is a public health problem. Is public health law the solution? Health Affairs Forefront [Internet]; 2026 Mar 16. Available from: https://www.healthaffairs.org/content/forefront/digital-addiction-public-health-problem-public-health-law-solution. Accessed April 19, 2026
- 21.↵GrabensteinH. What legal experts say about a major ‘bellwether trial’ over child social media addiction [Internet]; 2026. Available from: https://www.pbs.org/newshour/nation/what-to-know-about-a-trial-that-will-test-tech-giants-liability-for-child-social-media-addiction. Accessed May 16, 2026
- 22.↵KangCMacRTanE. Meta and YouTube found negligent in landmark social media addiction case. The New York Times [Internet]; 2026 Mar 25. Available from: https://www.nytimes.com/2026/03/25/technology/social-media-trial-verdict.html. Accessed May 16, 2026
- 23.↵JonesDN. Meta ordered to pay $375 million in New Mexico trial over child exploitation, user safety claims [Internet]; 2026. Available from: https://www.reuters.com/sustainability/boards-policy-regulation/jury-orders-meta-pay-375-mln-new-mexico-lawsuit-over-child-sexual-exploitation-2026-03-24/. Accessed May 16, 2026
- 24.↵HeadKR. Minds in crisis: How the AI revolution is impacting mental health. J Ment Health Clin Psychol. 2025; 9(3):34–44
- 25.↵HeadKR. Digital companions, real casualties: A commentary on rising AI-related mental health crises. Curr Res Psychiatry. 2026; 6(1):1–5
- 26.↵HillK. A teen was suicidal. ChatGPT was the friend he confided in. The New York Times [Internet]; 2025 Aug 26. Available from: https://www.nytimes.com/2025/08/26/technology/chatgpt-openai-suicide.html. Accessed May 16, 2026
- 27.↵OrtutayB. OpenAI faces 7 lawsuits claiming ChatGPT drove people to suicide, delusions [Internet]; 2026. Available from: https://apnews.com/article/openai-chatgpt-lawsuit-suicide-56e63e5538602ea39116f1904bf7cdc3. Accessed May 16, 2026
- 28.↵BlackK. New California lawsuits against OpenAI bring to 11 the number of claims linking chatbot to mental health harms [Internet]; 2026. Available from: https://www.law.com/therecorder/2026/01/26/new-california-lawsuits-against-openai-bring-to-11-the-number-of-claims-linking-chatbot-to-mental-health-harms/?slreturn=20260516034946. Accessed May 16, 2026
- 29.↵Texas v. Johnson, 491 U.S. 397 (1989)
- 30.↵Miles v. City Council of Augusta, GA., 710 F.2d 1542 (11th Cir. 1983)
- 31.↵BomboyS. Lawsuit analyzes First Amendment protection of AI chatbots in civil case [Internet]; 2025. Available from: https://constitutioncenter.org/blog/lawsuit-analyzes-first-amendment-protection-for-ai-chatbots-in-civil-case. Accessed April 19, 2026
- 32.↵Commonwealth v. Carter, 52 N.E.3d 1054 (Mass 2016)
- 33.↵GhossoubELandessJNewmanWJ. Commonwealth v. Carter and legal interpretations of facilitated suicide. J Am Acad Psychiatry Law. 2018 Dec; 46(4):521–31
- 34.↵PhillipsKM. Sticks and stones may break your bones, but words can also kill: Limiting criminal liability for words. U Ill L Rev. 2019:1741–70
- 35.↵MorrinHNichollsLLevinM. Artificial intelligence-associated delusions and large language models: Risks, mechanisms of delusion co-creation, and safeguarding strategies. Lancet Psychiatry. 2026; 13(6):522–30
- 36.↵CaldwellMRHoPA. A case of artificial-intelligence psychosis co-occurring with substance-induced psychosis. Prim Care Companion CNS Disord. 2025; 27(6):25cr04059
- 37.↵MooreJMehtaAAgnewW. Characterizing delusional spirals through human-LLM chat logs. Presented at: 2026 ACM Conference on Fairness, Accountability, and Transparency; 2026 June 25–28; Montreal, QC, Canada
- 38.↵DohnánySKurth-NelsonZSpensE. Technological folie à deux: Feedback loops between AI chatbots and mental illness. Nat Ment Health. 2026; 4(3):336–45
- 39.↵OslerL. Hallucinating with AI: Distributed delusions and “AI psychosis.” Philos Technol. 2026; 39(1):30
- 40.↵HudonAStipE. Delusional experiences emerging from AI chatbot interactions or “AI psychosis.” JMIR Ment Health. 2025; 12:e85799
- 41.↵DupréMH. Man who had managed mental illness effectively for years says ChatGPT sent him into hospitalization for psychosis [Internet]; 2026. Available from: https://futurism.com/artificial-intelligence/mental-illness-chatgpt-psychosis-lawsuit. Accessed April 23, 2026
- 42.
- 43.↵ShenMKYoonD.The dark addiction patterns of AI chatbot interfaces. Presented at: CHI EA ’25: Extended Abstracts of the CHI Conference on Human Factors in Computing Systems; 2025 April 26; Yokohama, Japan
- 44.↵RooseK. Can AI be blamed for a teen’s suicide? The New York Times [Internet]; 2024 Oct 23. Available from: https://www.nytimes.com/2024/10/23/technology/characterai-lawsuit-teen-suicide.html. Accessed April 20, 2026
- 45.↵JargonJ. Gemini said they could only be together if he killed himself. Soon, he was dead. The Wall Street Journal [Internet]; 2026 Mar 4. Available from: https://www.wsj.com/tech/ai/gemini-ai-wrongful-death-lawsuit-cc46c5f7. Accessed April 20, 2026
- 46.↵KerrD. Google faces lawsuit after Gemini chatbot allegedly instructed man to kill himself. The Guardian [Internet]; 2026 Mar 4. Available from: https://www.theguardian.com/technology/2026/mar/04/gemini-chatbot-google-jonathan-gavalas. Accessed April 20, 2026
- 47.↵KleeM. He had a mental breakdown talking to ChatGPT. Then police killed him. Rolling Stone [Internet]; 2025 Jun 22. Available from: https://www.rollingstone.com/culture/culture-features/chatgpt-obsession-mental-breaktown-alex-taylor-suicide-1235368941/. Accessed April 20, 2026
- 48.↵JoshiKGFriersonRLGunterTD. Shared psychotic disorder and criminal responsibility: A review and case report of folie à trois. J Am Acad Psychiatry Law. 2006 Dec; 34(4):511–7
- 49.↵NewmanWJHarbitMA. Folie à deux and the courts. J Am Acad Psychiatry Law. 2010 Sep; 38(3):369–75
- 50.↵KleeM. He was indicted for cyberstalking. His former friends tracked his ChatGPT meltdown. Rolling Stone [Internet]; 2026 Jan 14. Available from: https://www.rollingstone.com/culture/culture-features/chatgpt-ai-cyberstalking-social-media-1235496884/. Accessed April 23, 2026
- 51.↵BellanR. Stalking victim sues OpenAI, claims ChatGPT fueled her abuser’s delusions and ignored her warnings [Internet]; 2026. Available from: https://techcrunch.com/2026/04/10/stalking-victim-sues-openai-claims-chatgpt-fueled-her-abusers-delusions-and-ignored-her-warnings/. Accessed April 23, 2026
- 52.↵DupréMH. AI delusions are leading to domestic abuse, harassment, and stalking. Futurism [Internet]; 2026. Available from: https://futurism.com/artificial-intelligence/ai-abuse-harassment-stalking. Accessed April 23, 2026
- 53.↵SingletonTGerkenTMcMahonL. How a chatbot encouraged a man who wanted to kill the Queen. BBC [Internet]; 2023. Available from: https://www.bbc.com/news/technology-67012224. Accessed April 23, 2026
- 54.↵JargonJKesslerS. A troubled man, his chatbot and a murder-suicide in Old Greenwich. The Wall Street Journal [Internet]; 2025 Aug 28. Available from: https://www.wsj.com/tech/ai/chatgpt-ai-stein-erik-soelberg-murder-suicide-6b67dbfb. Accessed April 23, 2026
- 55.↵FollmanM. The chilling role of ChatGPT in mass shootings and other violence. Mother Jones [Internet]; 2026 Apr 10. Available from: https://www.motherjones.com/media/2026/04/chatgpt-tumbler-ridge-fsu-openai-chatbots-mass-shootings/. Accessed April 23, 2026
- 56.↵MeloyJRO’TooleME. The concept of leakage in threat assessment. Behav Sci & L. 2011 Jul-Aug; 29(4):513–27
- 57.↵MeloyJRRahmanT. Cognitive-affective drivers of fixation in threat assessment. Behav Sci & L. 2021 Apr; 39(2):170–89
- 58.↵LiorA. Tarasoff meets the AI age [Internet]; 2026. Available from: https://www.lawfaremedia.org/article/tarasoff-meets-the-ai-age. Accessed April 26, 2026
- 59.↵FollowsT. People are now marrying AI: Inside the rise of synthetic intimacy. Forbes [Internet]; 2025 Nov 15. Available from: https://www.forbes.com/sites/traceyfollows/2025/11/15/people-are-now-marrying-ai-inside-the-rise-of-synthetic-intimacy/. Accessed April 26, 2026
- 60.↵DimriA. The people who marry chatbots. The Atlantic [Internet]; 2026 Jan 2. Available from: https://www.theatlantic.com/ideas/2026/01/chatbot-marriage-ai-relationships-romance/685459/. Accessed April 26, 2026
- 61.↵SeinfeldS. The women who lost their AI boyfriends [Internet]; 2026. Available from: https://www.thefp.com/p/the-women-who-lost-their-ai-boyfriends. Accessed May 4, 2026
- 62.↵HorwitzJ. Meta’s flirty AI chatbot invited a retiree to New York. He never made it home [Internet]; 2025. Available from: https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/. Accessed May 5, 2026
- 63.↵FikeA. 75-year-old man asks wife for divorce after falling in love with AI chatbot [Internet]; 2025. Available from: https://www.vice.com/en/article/75-year-old-man-asks-wife-for-divorce-after-falling-in-love-with-ai-chatbot/. Accessed May 5, 2026
