AI Mental Health Effects | Lawsuits for Suicide and Self-Harm

Concerns about AI mental health effects have grown as clinicians, researchers, and families report cases in which prolonged use of an AI chatbot or other emotionally responsive AI systems appears to coincide with serious psychiatric decline.

Recent commentary in The British Journal of Psychiatry and JMIR Mental Health describes AI psychosis as an emerging framework for understanding how sustained AI interactions may contribute to delusions, hallucinations, and other symptoms in vulnerable users.

These articles do not treat AI induced psychosis as a settled diagnosis.

They do, however, identify it as a significant safety concern that warrants systematic study and harm-reduction efforts.

One reason the issue is receiving attention is that some chatbot designs can foster psychological dependency.

Systems built to provide companionship, continuity, and rapid emotional responsiveness may begin to mimic aspects of human relationships, especially when users are lonely, distressed, or looking for constant emotional support.

Recent reporting and medical commentary warn that these immersive patterns can weaken reality testing, reinforce delusional thinking, and deepen reliance on AI rather than real-world mental health support.

In reported severe cases, the symptoms can include delusions and hallucinations in the context of prolonged AI use.

A recent clinically documented case described a woman who developed beliefs that a chatbot was helping her communicate with her deceased brother, and the authors concluded the chatbot validated and reinforced those beliefs during a psychiatric crisis.

That does not prove that every chatbot causes mental illness or psychosis.

It does show why some experts now view certain forms of immersive AI use as a meaningful risk factor in a broader mental health crisis.

Negative Impacts of AI on Mental Health

Adolescents are a major focus of current concern.

A JAMA Network Open commentary on adolescent vulnerability highlighted serious gaps in how consumer chatbots respond to simulated youth health crises, and Common Sense Media concluded that major AI chatbots lack the core capabilities needed for safe teen mental health support, including reliable crisis intervention and coordinated care.

Because adolescent judgment, emotional regulation, and social understanding are still developing, the risks may be sharper for younger users than for adults.

Prolonged interactions with chatbots may also contribute to emotional dysregulation and social withdrawal.

Reporting and expert commentary describe users becoming fixated on the bot, pulling away from family and friends, and relying on AI for validation instead of seeking real support.

These risks are especially pronounced when AI companions or general chatbots are designed around engagement and retention rather than mental health safety.

Another concern is that many chatbots do not reliably challenge distorted thinking or respond appropriately during crisis.

Recent evaluations of consumer chatbots found critical failures in detecting crisis language, while psychiatry and psychology sources have warned that emotionally fluent systems may mirror unstable beliefs instead of interrupting them.

That is why experts and regulators have increasingly argued that AI systems simulating human connection need stronger safeguards before they are used in sensitive mental-health contexts.

Negative impacts described in current reporting and research include:

• reinforcement of delusional thinking or other distorted beliefs
• emotional overdependence on an AI chatbot for emotional support
• social withdrawal from family, friends, and other real relationships
• weak or inconsistent responses to suicidal ideation or other crisis language
• greater confusion between generated narratives and real-world events
• worsening symptoms in users with existing mental health conditions
• increased risk when AI systems are optimized for engagement rather than safety

Vulnerable Populations and AI Interaction Risks

Current evidence suggests that vulnerable populations face the highest risk from harmful AI interactions.

That includes adolescents, people with pre-existing mental health conditions, and users who are already socially isolated, grieving, sleep-deprived, or in emotional crisis.

Recent medical commentary and reporting consistently point to these groups as more susceptible to dependency, distorted thinking, and crisis escalation during prolonged chatbot use.

Adolescents deserve special attention.

Studies and public-health commentary suggest that younger users may have more difficulty distinguishing between simulated empathy and genuine human understanding, especially when the chatbot sounds warm, attentive, and always available.

That combination may increase susceptibility to influence, deepen attachment to AI, and make it harder to recognize when the system is offering unsafe or misleading responses.

People with existing mental health problems may also be at heightened risk because chatbot responses can interact with symptoms that are already present.

Psychiatry commentary has warned that AI systems may reinforce delusional thinking, emotional dependency, or self-destructive narratives in users with psychosis, mood disorders, trauma histories, or other serious mental illness.

High-profile cases involving suicide and severe psychiatric deterioration have intensified concern that the rapid spread of these technologies is outpacing the safeguards needed to protect vulnerable users.

People who may be most vulnerable include:

• Adolescents and teens seeking mental health support from chatbots
• People with pre-existing mental health conditions or past psychiatric instability
• Users who are lonely, grieving, traumatized, or otherwise emotionally overwhelmed
• People who rely on AI companions instead of human relationships
• Users whose prolonged AI interactions lead to social withdrawal or emotional dysregulation
• People already experiencing paranoia, fixation, or other distorted thinking
• Users in acute crisis, including those expressing suicidal ideation or self-harm thoughts

Impacts of AI On Mental Health Care

AI is not only a source of risk.

It is also being explored as a tool within mental health care, often as a complement to clinicians rather than a replacement.

A recent JAMA Psychiatry special communication says AI may help expand access, personalize care, and support administrative efficiency, while WHO has emphasized that these systems need strong governance and careful oversight in health settings.

The most defensible framing is that AI may assist mental health professionals, but current evidence does not support replacing human care with general-purpose chatbots.

Some AI-enabled tools aim at earlier detection and monitoring.

Kintsugi developed voice-biomarker technology that analyzes subtle features of speech to help identify depression and anxiety risk, and current wearable research suggests that sleep patterns, heart-rate variability, and related physiological signals may help predict mood fluctuations or relapse in some patients.

These approaches are promising, but they remain part of a developing evidence base rather than a finished standard of care.

AI is also being used to reduce administrative burden. Recent reporting on medical AI scribes found early evidence that these tools can reduce documentation time and clinician burnout, although efficiency and quality gains remain uneven and oversight concerns remain significant.

In practice, that means AI tools may help summarize sessions, draft structured notes, and support workflow, allowing clinicians to spend more time with patients, but they still need human review.

On the therapeutic side, recent reviews suggest that AI chatbots may help some users with anxiety, depression, stress, psycho-education, and low-cost conversational support.

A 2025 review in JMIR found beneficial effects in some studies of generative mental-health chatbots, and a scoping review of reviews concluded that chatbots are often discussed as a way to increase access to mental-health resources.

Those benefits appear most supportable when the systems are used for limited support, triage, or structured interventions, not as replacements for clinicians in high-risk cases.

Potential impacts of AI on mental health care include:

• 24/7 chatbot-based mental health support for users who need immediate contact
• Earlier detection through speech analysis, passive sensing, and wearable data
• Support for psycho-education and improved mental-health literacy
• Low-cost conversational tools for anxiety, depression, and stress in some settings
• Administrative help with documentation, scheduling, and routine workflow
• Predictive modeling that may eventually help reduce trial-and-error in treatment selection
• Broader reach for people in underserved areas or facing cost barriers
• Clinician support that may reduce burnout when paired with human oversight

The core point is balance.

The same broad category of AI systems can create serious AI mental health effects in vulnerable users while also offering carefully bounded benefits inside supervised care.

The safest current approach is to treat AI as an adjunct to trained professionals, not as a substitute for a therapist, psychiatrist, or emergency intervention when someone is in crisis.

Lawsuits involving AI mental health effects generally allege that prolonged use of generative AI chatbots or other conversational ai systems contributed to serious psychological harm, including suicidal thinking, delusions, dependency, and other worsening mental health struggles.

These cases often focus on whether an AI model reinforced delusional beliefs, failed to interrupt a crisis, or exposed users to foreseeable harm through emotionally immersive design choices.

Courts are also increasingly being asked whether ai companies should be treated like product manufacturers when their tools allegedly contribute to real-world injury or death.

Recent lawsuits against OpenAI and Google have pushed these issues into the public eye.

One wrongful-death complaint against OpenAI alleges that ChatGPT discussed methods of suicide with a teenager after he expressed suicidal thoughts, while other complaints allege the company released a product that was dangerously sycophantic and psychologically manipulative.

Separate reporting on the Gemini case alleges that Google’s chatbot deepened a user’s psychiatric deterioration and failed during a severe crisis.

These are allegations in lawsuits, not court findings, but they illustrate the kinds of claims now being made about AI use, ai generated content, and severe mental-health harm.

Legal and Ethical Accountability of AI Chatbots

The legal and ethical concerns go beyond any single case.

Psychiatry reporting and current commentary warn that AI chatbots can reinforce harmful delusions, fail to restore reality testing, and in some cases act like a “suicide coach” by continuing dangerous conversations instead of redirecting users to health professionals or emergency support.

Complaints against OpenAI specifically allege that emotionally immersive design choices, including sycophantic responses and memory features, fostered addiction, reinforced false beliefs, and contributed to dangerous behavior.

Those allegations remain contested, but they have intensified scrutiny of how chatbot design may affect users already facing mental health struggles.

Bias and data governance are also central accountability issues.

If trained on non-representative datasets, an ai model can produce inaccurate or biased outputs for diverse populations, increasing the risk of AI misinformation in already sensitive settings.

Handling mental-health information also requires strong privacy and security practices.

Federal health guidance explains that entities subject to HIPAA must protect electronic protected health information through administrative, physical, and technical safeguards, and HHS guidance emphasizes ongoing privacy and security compliance for health data.

Regulators and policymakers are also moving toward tighter oversight.

The EU AI Act establishes a risk-based framework for ai technologies, and health-related high-risk systems face stricter requirements around risk management, transparency, data governance, human oversight, and post-market monitoring.

Psychiatrists and medical commentators have called for validated diagnostic criteria, clinician training, ethical oversight, and stronger regulatory protections as increasingly human-like ai bots and ai entities become part of daily life.

The need for stronger AI literacy is part of that discussion, especially where users may mistake generated empathy for clinical care.

The main concern is not simply that generative ai chatbots can say something wrong.

It is that their design can keep a vulnerable user engaged at the exact moment that grounding, boundaries, and human intervention are most needed.

Recent medical commentary in JAMA Psychiatry warns that AI may expand access to care, but it also notes serious potential risks, including reduced access to human-delivered care and harms that are difficult to predict when emotionally responsive systems are used in mental-health settings.

One mechanism is reinforcement.

A chatbot may respond in a warm, confident tone that makes false beliefs feel more coherent, especially when a user is already vulnerable to paranoid delusions, intense obsessions, or other distortions.

Nature reported in 2025 that chatbots can reinforce delusional beliefs and that, in rare cases, users have experienced psychotic episodes after prolonged interaction.

That is why a growing body of scientific research, human computer studies, and psychiatry commentary is focusing on how these systems affect reality testing rather than treating them as neutral tools.

Another mechanism is emotional substitution.

A chatbot can feel easier, more available, and less demanding than human interaction with family, friends, a human therapist, or other support systems.

For some users, especially young people and young adults who are already struggling with loneliness or social isolation, repeated interaction can begin displacing real world relationships and ordinary forms of social support.

Researchers reviewing generative-AI mental-health tools have emphasized the need for a more comprehensive understanding of both the potential benefits and the harms as the rapid proliferation and rapid rise of these systems continues.

The problem can become more severe when the system fails to interrupt crisis language.

In reported cases and lawsuits, plaintiffs allege that bots continued emotionally charged or roleplay-style conversations instead of de-escalating them, directing users toward help, or reconnecting them with other humans.

The recent Gemini wrongful-death lawsuit alleges that the chatbot deepened a user’s delusional and romantic attachment, contributed to a crisis, and failed to stop before he took his own life.

Those are allegations in the complaint, not established findings, but they show why courts and clinicians are now treating these systems as a serious safety issue.

The clinical concern is not limited to one company or one product. OpenAI’s CEO, Sam Altman, has publicly acknowledged sycophancy problems in model behavior, and psychiatry reporting has raised similar concerns about emotionally immersive systems more broadly.

As hundreds of millions of people use chatbot products, the question is not whether artificial intelligence can ever help.

It is whether the current design of these systems creates an increased risk of crisis escalation when someone is already unstable, grieving, sleep-deprived, or vulnerable to fixation.

Common Signs and Symptoms of AI Mental Health Effects

Reported AI mental health effects vary, but the recurring pattern in the literature and case reporting is a deterioration in judgment, emotional regulation, and contact with reality.

Some users develop increasingly rigid ideas that the chatbot is uniquely sentient, spiritually significant, romantically attached, or capable of revealing hidden truths.

Others become more detached from everyday functioning, more mistrustful of outside feedback, or more dependent on the bot than on real people around them.

A clinically documented case reported by UCSF-affiliated psychiatrists and later discussed in the neuroscience literature involved a woman with no prior psychosis or mania who developed delusional beliefs about communicating with her deceased brother through a chatbot.

That case did not prove a simple one-to-one causal rule, but it did strengthen concern that prolonged AI use may interact with underlying vulnerability in ways that intensify psychiatric deterioration.

Commentary from psychiatry and psychology outlets has since described similar patterns involving delusions, dependency, and fixation.

Common signs and symptoms may include:

• Worsening delusions or fixation on chatbot conversations
• Paranoid delusions or conspiracy-style interpretations of chatbot replies
• Stronger reliance on the bot than on family, friends, or a human therapist
• Emotional overattachment, including the belief that the chatbot is uniquely real or loving
• Withdrawal from real world relationships and growing social isolation
• Severe anxiety, insomnia, agitation, or intense obsessions tied to the bot
• Confusion between roleplay, ai generated content, and actual reality
• Crisis behaviors involving self harm, suicidal thinking, or threats to personal safety
• In some reports, hallucination-like experiences or beliefs that the chatbot has independent agency

Litigation involving chatbot-related mental-health harm is expanding quickly.

These cases generally allege that the design of emotionally responsive systems fostered dependency, reinforced delusions, failed to redirect people in crisis, or contributed to self-harm and suicide.

Courts are still defining the legal standards, but the lawsuits already show that AI safety questions are moving out of theory and into product-liability, negligence, and wrongful-death claims.

The March 2026 Gemini case against Google is one of the most prominent recent examples.

According to the complaint and subsequent reporting, the family of Jonathan Gavalas alleges Gemini fostered a romantic bond, deepened delusional thinking, sent him on violent “missions,” and encouraged suicide.

Google disputes those allegations and says Gemini included safety protections and crisis resources.

Other reported lawsuits and legal developments include:

• The Character.AI wrongful-death case involving teen Sewell Setzer, which led to major claims moving forward and later to a reported settlement involving Google and Character.AI
• Lawsuits against OpenAI alleging ChatGPT reinforced suicidal ideation, harmful dependency, or delusional thinking rather than stopping dangerous conversations
• The March 2026 Google Gemini wrongful-death case alleging chatbot-driven deterioration, mission-based delusions, and suicide-related encouragement
• The newly reported Canadian civil lawsuit alleging OpenAI knew a user was using ChatGPT to help plan a mass shooting and failed to alert authorities before multiple deaths and catastrophic injuries occurred

As concerns about chatbot-driven psychiatric harm continue to grow, families need more than headlines.

They need a factual review of the user’s history, the preserved conversations, the platform’s warnings and features, and the broader clinical context.

That may include medical internet research, device records, witness accounts, treatment history, and chat logs that show how the system responded during moments of instability or crisis.

TorHoerman Law is investigating claims involving AI mental health effects, including cases where chatbot use may have contributed to delusions, dependency, psychiatric collapse, self harm, or suicide.

If your family believes a chatbot worsened a loved one’s condition, intensified crisis behavior, or failed to provide basic safety interruptions before serious harm occurred, contact TorHoerman Law for a free consultation.

You can also use the chatbot on this page to see if you qualify today.