A Chatbot Warning Changed What Students Said They'd Do

experiment · Frontiers in psychology · la publicación, 1 sep 2026 · gratis

Un mensaje breve antes de escribir: así cambian lo que usted decide

Un estudio con estudiantes en China sugiere que dos avisos visibles modifican lo que la gente dice que haría. No probó qué pasa después.

Versión breve · la versión detallada sigue, unos 8 min

Pregunte a weeklyAI

Pregúnteme por este estudio: a quiénes se estudió, qué encontró y qué no dice.

Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.

El estudio, de un vistazo
Quiénes: Estudiantes universitarios
Cuántos: 768 estudiantes universitarios, repartidos al azar en cuatro grupos de 192
Dónde: China
Cuándo: No lo dice el pasaje
Tipo de estudio: experimento
Quién lo hizo: Investigadores chinos; estudio aprobado por la Universidad de Ciencias Políticas y Derecho de Xinjiang, en Tumxuk, China
El límite que importa: Fueron escenarios hipotéticos leídos una sola vez, no conversaciones reales con un chatbot.

Confianza mejor equilibrada en cada grupo de avisos

Sin ningún aviso3.34puntos

Solo aviso de privacidad3.55puntos

Solo advertencia de límites3.49puntos

Ambos avisos3.98puntos

Puntuación media de confianza mejor equilibrada en cada uno de los cuatro grupos del experimento; son respuestas declaradas ante un escenario escrito, no conducta real.

Efecto de cada aviso sobre lo que los estudiantes dijeron

Aviso de privacidadfrente aSin aviso de privacidad

Aumentó la sensación de que la información estaría protegida

Advertencia de límites profesionalesfrente aSin advertencia de límites

Aumentó la comprensión de que el chatbot tiene un techo profesional

Advertencia de límitesfrente aAviso de privacidad

Solo la advertencia de límites se asoció con menor disposición a depender del chatbot y mayor intención de buscar ayuda profesional

Ver · 0:58

Muchos estudiantes universitarios ya preguntan a los chatbots de inteligencia artificial por el estrés, el sueño o el malestar emocional. Son herramientas fáciles de alcanzar y que no juzgan. Pero plantean dudas: ¿qué pasa con lo que uno cuenta? ¿Hasta dónde puede llegar esa ayuda?

Un grupo de investigadores chinos quiso saber si dos mensajes breves en la pantalla del chatbot cambian cómo la gente lo evalúa. Uno es la garantía de privacidad: una frase que explica cómo se manejan los datos. El otro es la advertencia de límites: una frase que aclara que el chatbot no sustituye a un profesional.

Hicieron un experimento con 768 estudiantes universitarios, repartidos al azar en cuatro grupos de 192. Cada grupo vio una versión distinta del mismo escenario: sin los dos avisos, solo con la garantía de privacidad, solo con la advertencia de límites, o con ambos. Luego respondieron preguntas sobre lo que pensaban y lo que harían.

La garantía de privacidad aumentó la sensación de que la información estaba protegida. La advertencia de límites aumentó la conciencia de que el chatbot tiene fronteras, redujo el riesgo de apoyarse demasiado en él y aumentó la intención de buscar ayuda profesional. La confianza medida —ni ciega ni desconfiada— fue más alta cuando aparecían los dos mensajes juntos.

Esto no fue una conversación real con un chatbot. Fueron escenarios hipotéticos leídos una sola vez. Lo que se midió fueron intenciones y percepciones declaradas, no lo que la gente hace después. Las intenciones no siempre predicen la conducta. Nadie puede concluir que estos avisos logren que alguien busque ayuda de verdad.

Tampoco se puede trasladar el resultado a otros países. Los participantes eran estudiantes universitarios en China, un solo contexto nacional. El estigma, los servicios de consejería y las reglas de protección de datos pueden variar de un país a otro. El estudio no evaluó si el chatbot diagnosticaba, trataba ni atendía crisis, y no midió ningún beneficio clínico.

Si una universidad o una empresa le ofrece un chatbot de apoyo emocional, usted puede pedir que los avisos estén a la vista antes de escribir nada: qué datos se recogen, quién los ve, cuánto se guardan, y una ruta clara hacia atención real cuando haga falta.

Antes de contarle algo íntimo a una máquina, pregunte: ¿me dice esta herramienta, con claridad, qué hará con lo que escribo y en qué momento debo llamar a una persona?

Qué significa para usted

Cuando usted vea un aviso de privacidad antes de escribir, lo que cambia es su sensación de que los datos están protegidos, según este estudio. El aviso de límites es el que se relacionó con buscar ayuda profesional y con apoyarse menos en la máquina. Pero eso aún no prueba qué hará usted después: son intenciones declaradas en un escenario, no conducta real.

Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264

Quién pagó: Los autores declararon que no recibieron apoyo financiero para este trabajo ni para su publicación, y no declararon relaciones comerciales o financieras que pudieran constituir un conflicto de interés.

No tome esto como consejo médico profesional.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1513 palabras · unos 8 minLeerla →Cerrar

Los mensajes antes de escribir: la advertencia de límites redujo la dependencia y aumentó la intención de buscar ayuda profesional

Un experimento con 768 estudiantes universitarios en China encontró que la advertencia de límites aumentó la comprensión de que el chatbot tiene un techo profesional, redujo la disposición a depender de él y fortaleció la intención de buscar ayuda profesional, mientras que el aviso de privacidad aumentó la sensación de que la información estaría protegida.

Así podría verse · ilustración generada por weeklyAI.watch, no es una fotografía

Antes de que alguien escriba en un chatbot de inteligencia artificial que le duele algo, la aplicación puede mostrarle dos mensajes distintos: uno que explica cómo se manejará su información, y otro que aclara que la máquina no sustituye a un profesional de salud mental¹². Un equipo de investigadores chinos encontró que la advertencia de límites aumentó la comprensión de que el chatbot tiene un techo profesional, redujo la disposición a depender de él y fortaleció la intención de buscar ayuda profesional, mientras que el aviso de privacidad aumentó la sensación de que la información estaría protegida. Para averiguarlo, reunieron a 768 estudiantes universitarios y los repartieron al azar en cuatro grupos iguales de 192 personas cada uno: un grupo vio un chatbot sin ningún aviso; otro, solo el aviso de privacidad; otro, solo la advertencia de límites profesionales; y el último, los dos mensajes juntos. Cada participante leyó una sola situación imaginaria y luego respondió un cuestionario. Nadie habló con un chatbot real.

Los resultados fueron claros en dos puntos. El aviso de privacidad aumentó la sensación de que la información estaría protegida; la advertencia de límites aumentó la comprensión de que el chatbot tiene un techo profesional³. El grupo que vio los dos mensajes fue el que mostró la confianza mejor equilibrada: la puntuación media más alta entre los cuatro grupos⁴. Y fue la advertencia de límites, no el aviso de privacidad, la que se asoció con una menor disposición a depender del chatbot cuando haría falta ayuda profesional, y con una mayor intención de buscar esa ayuda⁵. Los autores concluyen que los dos mensajes funcionan mejor juntos: uno atiende el miedo a contar algo íntimo, el otro aclara hasta dónde llega la herramienta⁶.

Conviene ser preciso sobre lo que esto significa. El estudio no midió si alguien pidió ayuda de verdad, ni si un chatbot resolvió una crisis, ni si la herramienta sirve clínicamente⁷. Midió lo que estudiantes chinos dijeron que sentirían y que harían ante una situación escrita. Las autoras y autores lo advierten: las intenciones declaradas no son comportamiento, y las respuestas pueden variar de un país a otro por diferencias en el estigma, los servicios de consejería y las normas sobre datos⁸. Los autores declararon que no recibieron apoyo financiero para este trabajo ni para su publicación, y que no tenían relaciones comerciales que pudieran considerarse un conflicto de interés. Fue aprobado por la oficina de investigación de la Universidad de Ciencias Políticas y Derecho de Xinjiang, en Tumxuk, China.

Vale la pena entender qué es esta tecnología, porque no es la de hace una década. Los chatbots anteriores seguían guiones fijos; los modelos de lenguaje actuales generan respuestas fluidas y adaptadas a casi cualquier tema⁹. Esa fluidez es justamente lo que engaña: puede hacer que la máquina parezca más competente de lo que es y ocultar que inventa datos o que nadie responde profesionalmente por lo que dice⁹. Por eso los propios autores insisten en que un chatbot puede servir para informarse y orientarse, pero no debe tratarse como herramienta de diagnóstico, psicoterapia, intervención en crisis ni decisión médica¹⁰.

La tecnología ya está en manos de los estudiantes. La usan para buscar información sobre salud mental, pedir consuelo, explorar formas de sobrellevar el malestar e identificar dónde pedir apoyo¹¹. Y hay un problema de fondo que el estudio recoge: trabajos recientes sugieren que quienes conversan con chatbots de propósito general sobre temas emocionales pueden malentender qué protección real tienen sus confesiones¹². De ahí que los dos avisos probados sean tan concretos: el de privacidad explica qué se recoge, qué se protege y qué no se usará para otros fines; el de límites dice que la herramienta no reemplaza el diagnóstico, la psicoterapia, la atención de crisis ni el tratamiento médico, y que ante un malestar grave o urgente hay que acudir a profesionales o servicios de emergencia¹³¹⁴.

Para dimensionar el asunto, miremos lo que ocurre en otros sistemas de salud. Una encuesta disponible a los 53 Estados miembros de la Región Europea de la Organización Mundial de la Salud, realizada entre junio de 2024 y marzo de 2025, encontró que de los 50 países que respondieron, apenas el 8 % tenía una estrategia de inteligencia artificial específica para salud y el 14 % estaba desarrollándola, según el resumen de ese estudio; solo pudimos leer el resumen, el texto completo está detrás de una suscripción¹⁵¹⁶. Menos del 10 % había desarrollado estándares de responsabilidad para la inteligencia artificial, y solo 14 países habían emitido guías sobre sus implicaciones éticas en salud¹⁷¹⁸. La conclusión de ese trabajo es que la gobernanza de la inteligencia artificial en salud está subdesarrollada en esa región¹⁹.

Otros trabajos que leímos apuntan en la misma dirección. En cirugía, por ejemplo, los sistemas de inteligencia artificial ya se usan para apoyar decisiones y guiar procedimientos, pero introducen riesgos técnicos, humanos, legales y éticos que los marcos regulatorios actuales no alcanzan a cubrir, según el resumen de un documento de la Sociedad Estadounidense de Cirugía Gastrointestinal y Endoscópica; solo pudimos leer el resumen²⁰²¹²². Y en salud mental, un análisis de 33 sistemas encontró que los riesgos más frecuentes incluyen respuestas dañinas o inseguras, fallas en la respuesta a crisis, exposición de información personal sensible y poca transparencia; el 78,8 % de esos estudios fue calificado de alto riesgo por la debilidad de sus pruebas de seguridad, según su resumen²³²⁴.

Así lo leemos nosotros. Cuando alguien le cuenta algo íntimo a una máquina que responde con soltura, puede quedarse con la sensación de que ya fue escuchado y bajar la urgencia de contárselo a una persona real. El estudio sugiere que una advertencia clara de límites aumenta la intención declarada de buscar ayuda profesional, pero en la vida diaria muchos usuarios podrían sentir que con la conversación ya hicieron lo suficiente y postergar la consulta. Sabríamos que nos equivocamos si viéramos que esa intención se traduce en llamadas reales a servicios de consejería y que quienes más usan el chatbot no tardan más en buscar apoyo humano. Mientras tanto, si usted o alguien de su casa conversa con un chatbot sobre angustia, vale la pena preguntarse después si esa charla reemplazó o solo acompañó el paso de hablar con una persona calificada. Una regla simple: la conversación con la máquina no cuenta como consulta.

También leemos que la confianza se construye cuando lo que se dice coincide con lo que se hace. Los avisos de privacidad y de límites solo producirán una confianza bien calibrada si la empresa demuestra con acciones concretas lo que anuncia: qué guarda, quién lo ve y qué hace ante una crisis. Si el aviso es solo texto y las prácticas no coinciden, el efecto se debilita o se convierte en desconfianza. Antes de confiar en un chatbot de salud mental, busque su política de privacidad y verifique si dice qué datos guarda, por cuánto tiempo y quién puede verlos. Si no la encuentra o es ambigua, trátelo como una señal de alerta.

Y una distinción que ayuda a decidir a quién pedirle cuentas: cuando una aplicación le advierte que no reemplaza a un profesional, pregúntese si esa advertencia le aclara el papel real de la herramienta o solo protege a la empresa. Una advertencia honesta, que reconoce lo que el chatbot no puede hacer, probablemente se perciba como responsabilidad y no como una excusa para devaluar la herramienta. Si suena a descargo legal escondido, el usuario podría ignorarla o desconfiar de toda la aplicación.

Los propios autores señalan qué faltaría para dar el paso siguiente. Habría que probar estos avisos en interfaces reales de chatbot y medir comportamientos, no solo intenciones: por ejemplo, si la persona hace clic para pedir ayuda o si efectivamente usa un servicio de referencia²⁵. También proponen que un chatbot responsable declare desde el inicio que no puede diagnosticar, dar psicoterapia, reemplazar a consejeros o médicos, ni atender emergencias, y que repita esa advertencia cuando alguien menciona malestar severo, autolesión, daño a terceros, decisiones médicas o deterioro prolongado²⁶.

Aquí está lo que esto hace posible para usted. La próxima vez que una aplicación de salud mental le muestre un aviso antes de empezar, no lo salte: léalo como lo que es, una declaración sobre los límites de la herramienta y sobre quién responde por sus datos. Pregunte qué guarda, por cuánto tiempo y quién puede verlo. Y si usted o alguien de su familia usa un chatbot para hablar de angustia, trate esa conversación como un primer paso informativo, nunca como el paso final. ¿Qué le diría hoy a la aplicación que su familiar usa para saber si cumple lo que promete?

De dónde sale cada dato de contexto, y cuánto leímos de cada documento

Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Privacy assurance refers to a clear statement that informs users how mental health-related information is handled, protected, and limited in use."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Professional-boundary warning refers to a clear statement that the chatbot is not a substitute for qualified mental health care, clinical diagnosis, psychotherapy, crisis intervention, or emergency services."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Privacy assurance increased perceived privacy protection, F (1, 764) = 159.30, p < 0.001, ηp 2 = 0.172, d = 0.91. Professional-boundary warning increased boundary awareness, F (1, 764) = 176.40, p < 0.001, ηp 2 = 0.188, d = 0.96."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Calibrated trust was highest in the both-cues condition (M = 3.98, 95% CI [3.89, 4.07]) compared with privacy assurance only (M = 3.55, 95% CI [3.46, 3.64]), boundary warning only (M = 3.49, 95% CI [3.40, 3.58]), and no cues (M = 3.34, 95% CI [3.24, 3.44])."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Boundary awareness was associated with calibrated trust, β = 0.25, p < 0.001, lower overreliance risk, β = −0.35, p < 0.001, and stronger professional help-seeking intention, β = 0.35, p < 0.001."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The interaction result further suggests that the two messages work best together. A privacy assurance alone may make the tool easier to approach, but it does not by itself clarify when human care is needed. A boundary warning alone may limit inappropriate reliance, but it does not fully address disclosure concerns."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Because this study used hypothetical vignette scenarios rather than actual chatbot interactions, the findings should be interpreted as evidence about students' safety-related evaluations and intentions, not as evidence of real-world chatbot use or clinical help-seeking behavior."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The sample consisted of college students in one national context. The findings should be interpreted within the cultural and institutional context in which the study was conducted. Students' responses to privacy messages, professional-boundary warnings, and help-seeking recommendations may vary across countries"
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Unlike earlier rule-based or scripted conversational agents, large language models can generate fluent and context-sensitive responses across a very wide range of topics. Fluency may increase perceived competence, but it may also obscure uncertainty, hallucinated information, or the lack of professional accountability."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "In mental health-related settings, this creates a delicate tension: AI chatbots may serve as sources of general information and support navigation, but they should not be treated as diagnostic, psychotherapeutic, crisis-intervention, or medical decision-making tools"
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Generative artificial intelligence (AI) chatbots are increasingly used by students to look up mental health information, seek reassurance, explore coping strategies, and identify possible sources of support."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Recent work on general-purpose LLM chatbots for mental health also suggests that users may misunderstand the privacy and regulatory protections attached to emotionally sensitive disclosures."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "In the privacy-assurance condition, the scenario added a statement explaining that the chatbot would minimize the collection of personal information, protect mental health-related data, avoid using identifiable information for unrelated purposes, and clearly inform users how their information would be handled."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "In the professional-boundary warning condition, the scenario added a statement explaining that the chatbot was not a substitute for diagnosis, psychotherapy, crisis intervention, or medical treatment, and that users should contact qualified professionals or emergency services when distress is severe or urgent."
Adib K, Letchford N, Dunning HE, Salama N, Tolias Y, De Barros J, et al. (2026). Governance of artificial intelligence for health systems, WHO European Region. Bulletin of the World Health Organization. 10.2471/blt.25.294978 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The survey was available to all 53 Member States between June 2024 and March 2025."
Adib K, Letchford N, Dunning HE, Salama N, Tolias Y, De Barros J, et al. (2026). Governance of artificial intelligence for health systems, WHO European Region. Bulletin of the World Health Organization. 10.2471/blt.25.294978 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Of the 50 Member States responding to the survey, 8% (4/50) have a health-specific AI strategy and 14% (7) are developing one."
Adib K, Letchford N, Dunning HE, Salama N, Tolias Y, De Barros J, et al. (2026). Governance of artificial intelligence for health systems, WHO European Region. Bulletin of the World Health Organization. 10.2471/blt.25.294978 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Only 14 Member States have issued guidelines to address the ethical implications of using AI in health or across sectors."
Adib K, Letchford N, Dunning HE, Salama N, Tolias Y, De Barros J, et al. (2026). Governance of artificial intelligence for health systems, WHO European Region. Bulletin of the World Health Organization. 10.2471/blt.25.294978 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Less than 10% (4) of Member States have developed liability standards for AI or guidance on the application of existing liability standards."
Adib K, Letchford N, Dunning HE, Salama N, Tolias Y, De Barros J, et al. (2026). Governance of artificial intelligence for health systems, WHO European Region. Bulletin of the World Health Organization. 10.2471/blt.25.294978 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Conclusion In the WHO European Region, governance of AI in health care is underdeveloped."
Hashimoto DA, Marwaha JS, Lee SA, Schwaitzberg S, Duffourc MN. (2026). Risk and liability in the deployment of AI systems for surgery: a SAGES white paper. Surgical Endoscopy. 10.1007/s00464-026-12881-8 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Artificial intelligence (AI) is increasingly utilized in surgical care for decision support, operative planning, intraoperative guidance, and autonomous functions."
Hashimoto DA, Marwaha JS, Lee SA, Schwaitzberg S, Duffourc MN. (2026). Risk and liability in the deployment of AI systems for surgery: a SAGES white paper. Surgical Endoscopy. 10.1007/s00464-026-12881-8 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "While these systems can enhance efficiency and clinical performance, they also introduce risks related to technology, human factors, legal issues, and ethics."
Hashimoto DA, Marwaha JS, Lee SA, Schwaitzberg S, Duffourc MN. (2026). Risk and liability in the deployment of AI systems for surgery: a SAGES white paper. Surgical Endoscopy. 10.1007/s00464-026-12881-8 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Current regulatory and legal frameworks are not fully equipped to address the challenges of AI-assisted surgery."
Sawesi S, Sabbineni H, Shagamreddy R, Rashrash B. (2026). Cybersecurity and Privacy Risks of Generative AI Mental-Health Chatbots: A Systematic Review and Regulatory Framework. Journal of Multidisciplinary Healthcare. 10.2147/jmdh.s581251 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Key risks included harmful or unsafe outputs, failures in crisis response, exposure of sensitive personal information, and limited transparency."
Sawesi S, Sabbineni H, Shagamreddy R, Rashrash B. (2026). Cybersecurity and Privacy Risks of Generative AI Mental-Health Chatbots: A Systematic Review and Regulatory Framework. Journal of Multidisciplinary Healthcare. 10.2147/jmdh.s581251 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Critically, 78.8% of studies (26/33) were rated high risk for cybersecurity evaluation rigor, indicating that formal adversarial testing and structured threat modeling remain rare."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Future work should test similar safety cues in functioning chatbot interfaces and examine behavioral outcomes such as click-through to help resources or actual referral use."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A responsible chatbot should clearly state at onboarding that it cannot diagnose mental disorders, provide psychotherapy, replace counselors or physicians, or handle emergencies. Boundary messages should also appear again when users mention severe distress, self-harm, harm to others, medical decisions, or prolonged functional impairment."

No tome esto como consejo médico profesional.

Los hallazgos de otros estudios que aquí se mencionan los conocemos por este documento, que fue el que leímos; no abrimos cada uno de esos estudios.

experiment · Frontiers in psychology · the paper, 1 Sep 2026 · free

A Chatbot Warning Changed What Students Said They'd Do

In a test of 768 college students in China, messages about privacy and about the chatbot's limits shifted stated intentions — but nothing real was tested.

Short version · the longer version follows, about 6 min

Ask weeklyAI

Ask me about this study: who was studied, what it found, and what it does not say.

Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.

The study at a glance
Who: college students
How many: 768
Where: China
When: not stated in the passages
Kind of study: experiment
Who did it: five institutions in China, including hospitals and a middle school
The limit that matters: Students read a made-up scenario; no one talked to a real chatbot.

How balanced students' trust in the chatbot was, by which messages they saw

both messages3.98points on a five-point scale

privacy message only3.55points on a five-point scale

boundary message only3.49points on a five-point scale

neither message3.34points on a five-point scale

Average scores for balanced trust — seeing the chatbot as useful but limited — among students who read each made-up scenario; these are stated evaluations, not what students did with a real chatbot.

What each message did on its own

students who saw the privacy messageagainststudents who did not

felt more protected

students who saw the boundary messageagainststudents who did not

understood the chatbot's limits better

The boundary note, in particular, was tied to students saying they would rely on the chatbot less inappropriately and would more strongly intend to seek professional help when needed.
weeklyAI's reading

Watch · 0:47

A student types into a generative AI chatbot: stressed, not sleeping, not sure who to talk to. The chatbot answers right away. Before that first message, though, the screen may show something else — a note about what happens to what you type, or a note saying the chatbot cannot replace a counselor.

Researchers at five different institutions in China, including hospitals and a middle school, wanted to know whether those two notes change how students judge the chatbot.

They ran a randomized experiment with 768 college students. Each was assigned by chance to one of four made-up chatbot scenarios: no note at all, a privacy assurance only, a professional-boundary warning only, or both. The privacy note said personal information would be minimized and protected and that users would be told how their data is handled. The boundary note said the chatbot was not a substitute for diagnosis, psychotherapy, crisis intervention or medical treatment.

The students who saw the privacy note reported feeling more protected. The students who saw the boundary note showed greater awareness of the chatbot's limits. Trust that was balanced — seeing the tool as useful but limited — was highest when both notes appeared.

The boundary note, in particular, was tied to students saying they would rely on the chatbot less inappropriately and would more strongly intend to seek professional help when needed.

But here is the catch. This was a hypothetical scenario, not a working chatbot. Nobody actually chatted with anything. The study measured what students said they felt and intended, not what they or a chatbot would really do. Stated intentions do not perfectly predict behavior, so these messages are not proven to change real help-seeking.

The students were also all college students in one country, China. Students' responses may vary across countries because of differences in mental health stigma, campus counseling systems, data-protection norms, trust in institutions, and familiarity with AI tools, so the results may not carry over.

What the study does suggest is a question worth asking of any chatbot maker, university or clinic near you: before you type anything sensitive, does the screen tell you plainly what happens to your words — and does it tell you, just as plainly, when to stop and call a real person?

What this means for you

For now, this is a question to ask rather than a rule to follow: when a chatbot near you asks for something sensitive, look for a plain note about what happens to your words and one about when a real professional is the right person to ask. Nothing here proves these notes change what anyone actually does.

Who paid: The authors declared that no financial support was received for this work and/or its publication, and no commercial or financial relationships were declared as a potential conflict of interest.

Do not take this as professional medical advice.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1142 words · about 6 minRead it →Close

Telling students a mental health chatbot has limits raised their stated intent to seek professional help

A randomized experiment with 768 college students found that two short messages — one about privacy, one about limits — shifted what students said they would do, not what they actually did.

How it could look · illustration generated by weeklyAI.watch, not a photograph

Two brief messages shown before a mental health chatbot conversation changed how 768 Chinese college students judged the tool, according to a randomized vignette experiment published in Frontiers in Psychology. One message explained how mental health information would be handled, protected and limited in use¹. The other stated that the chatbot was not a substitute for qualified mental health care, clinical diagnosis, psychotherapy, crisis intervention or emergency services². Students were randomly assigned to one of four scenarios: neither message, privacy assurance alone, boundary warning alone, or both. The privacy message raised how well protected students felt their information was; the boundary message raised how clearly students understood the chatbot's limits³. The boundary message was also linked to lower willingness to lean on the chatbot when professional help would be more appropriate, and to a stronger stated intention to seek professional help⁴. Calibrated trust — confidence in the tool paired with awareness of its limits — was highest when both messages appeared together⁵.

The study cannot tell us what happens in a real conversation. Participants read a hypothetical scenario; they did not talk to a working chatbot, and the researchers say plainly that the findings are evidence about students' evaluations and intentions, not about real-world chatbot use or actual help-seeking behavior⁶. The outcomes were self-reported intentions and risk perceptions, and intentions do not perfectly predict behavior. The 768 students were enrolled at colleges in one country, China, and the authors caution that responses to privacy messages, boundary warnings and help-seeking recommendations may vary across countries⁷. The study was not designed to test whether a chatbot helps anyone clinically; no participant received diagnosis, psychotherapy or crisis support from the system, and crisis intervention was not tested. The authors report no financial support for the work and no commercial or financial relationships that could be a conflict of interest; the protocol was reviewed by the research office of Xinjiang University of Political Science and Law in Tumxuk, China.

The technology in question is not the scripted chatbot of a few years ago. Large language models generate fluent, context-sensitive answers across a very wide range of topics, and that fluency can make them seem more competent than they are — it can also obscure uncertainty, invented information and the absence of professional accountability⁸. Students increasingly use these tools to look up mental health information, seek reassurance, explore coping strategies and find sources of support⁹. The article describes a study in which a chatbot may serve as a source of general information and a guide toward support, but it should not be treated as a diagnostic, psychotherapeutic, crisis-intervention or medical decision-making tool¹⁰.

Here is the mechanism, step by step, in the study's own terms. A privacy assurance is a statement about what happens to what you type: what is collected, what is protected, what will not be done with identifiable information, how you will be told¹¹. A professional-boundary warning is a statement about what the tool is not: not a substitute for diagnosis, psychotherapy, crisis intervention or medical treatment, with a direction to contact qualified professionals or emergency services when distress is severe or urgent¹². The study's reasoning is that these two messages address two different risks — whether it is safe to disclose, and how far to rely — and that neither alone covers both¹³. Earlier work on general-purpose chatbots for mental health suggests users may misunderstand the privacy and regulatory protections attached to emotionally sensitive disclosures¹⁴.

To picture the size of the effect, the researchers report group averages on a five-point scale. Calibrated trust was highest when both messages were present, at an average of 3.98, compared with 3.55 with privacy assurance alone, 3.49 with the boundary warning alone, and 3.34 with neither⁵. The pattern the authors draw from this is that the two messages work best together: privacy assurance alone may make the tool easier to approach without clarifying when human care is needed, and a boundary warning alone may limit inappropriate reliance without fully addressing disclosure concerns¹³.

What would have to happen next is more concrete than another survey. The authors call for testing similar safety cues inside working chatbot interfaces and measuring behavioral outcomes such as whether users click through to help resources or actually use a referral¹⁵. They also say a responsible chatbot should state at the outset that it cannot diagnose mental disorders, provide psychotherapy, replace counselors or physicians, or handle emergencies, and should repeat that message when a user mentions severe distress, self-harm, harm to others, medical decisions or prolonged functional impairment¹⁶. Until such tests are run, no one can say these messages change what people do.

Here is how we read it. Two sentences on a screen are being asked to do the work of a relationship: to tell you who else may see what you type, and to tell you when the screen has reached its limit. Read that way, a privacy message is less a technical safeguard than a rule agreed in advance — and once a rule is agreed, a later surprise about stored or shared conversations lands differently than it would have without the promise. A boundary message, meanwhile, may work less by adding information than by removing the appearance of authority that fluent answers carry. We would expect the combination to matter most in a household where no one has yet sat with a counselor, and where a confident-sounding answer is the only answer available at two in the morning. We could be wrong in a way you can watch for: if students who read a clear privacy promise turn out to be no more willing to disclose than those who read nothing, or if a clear statement of limits does nothing to reduce the tendency to follow chatbot advice without checking it, the pattern falls apart.

What you can do with this does not depend on the study being right. Before you type anything sensitive into a chatbot, ask who else can see it, how long it is kept, and whether you can delete it — and if the answers are vague, treat the vagueness as the answer. Notice whether you are treating a confident answer as a qualified one; when a chatbot sounds sure about your mental health, ask what it is actually in a position to know. And if you are the person a younger family member comes to, the useful question is not whether the chatbot was helpful but whether it was the last thing they consulted. That is a question you can ask tonight, and it costs nothing to ask.

Where each piece of context comes from, and how much of it we read

Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Privacy assurance refers to a clear statement that informs users how mental health-related information is handled, protected, and limited in use."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Professional-boundary warning refers to a clear statement that the chatbot is not a substitute for qualified mental health care, clinical diagnosis, psychotherapy, crisis intervention, or emergency services."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Privacy assurance increased perceived privacy protection, F (1, 764) = 159.30, p < 0.001, ηp 2 = 0.172, d = 0.91. Professional-boundary warning increased boundary awareness, F (1, 764) = 176.40, p < 0.001, ηp 2 = 0.188, d = 0.96."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Boundary awareness was associated with calibrated trust, β = 0.25, p < 0.001, lower overreliance risk, β = −0.35, p < 0.001, and stronger professional help-seeking intention, β = 0.35, p < 0.001."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Calibrated trust was highest in the both-cues condition (M = 3.98, 95% CI [3.89, 4.07]) compared with privacy assurance only (M = 3.55, 95% CI [3.46, 3.64]), boundary warning only (M = 3.49, 95% CI [3.40, 3.58]), and no cues (M = 3.34, 95% CI [3.24, 3.44])."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Because this study used hypothetical vignette scenarios rather than actual chatbot interactions, the findings should be interpreted as evidence about students' safety-related evaluations and intentions, not as evidence of real-world chatbot use or clinical help-seeking behavior."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "The sample consisted of college students in one national context. The findings should be interpreted within the cultural and institutional context in which the study was conducted. Students' responses to privacy messages, professional-boundary warnings, and help-seeking recommendations may vary across countries"
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Unlike earlier rule-based or scripted conversational agents, large language models can generate fluent and context-sensitive responses across a very wide range of topics. Fluency may increase perceived competence, but it may also obscure uncertainty, hallucinated information, or the lack of professional accountability."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Generative artificial intelligence (AI) chatbots are increasingly used by students to look up mental health information, seek reassurance, explore coping strategies, and identify possible sources of support."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "In mental health-related settings, this creates a delicate tension: AI chatbots may serve as sources of general information and support navigation, but they should not be treated as diagnostic, psychotherapeutic, crisis-intervention, or medical decision-making tools"
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "In the privacy-assurance condition, the scenario added a statement explaining that the chatbot would minimize the collection of personal information, protect mental health-related data, avoid using identifiable information for unrelated purposes, and clearly inform users how their information would be handled."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "In the professional-boundary warning condition, the scenario added a statement explaining that the chatbot was not a substitute for diagnosis, psychotherapy, crisis intervention, or medical treatment, and that users should contact qualified professionals or emergency services when distress is severe or urgent."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "The interaction result further suggests that the two messages work best together. A privacy assurance alone may make the tool easier to approach, but it does not by itself clarify when human care is needed. A boundary warning alone may limit inappropriate reliance, but it does not fully address disclosure concerns."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Recent work on general-purpose LLM chatbots for mental health also suggests that users may misunderstand the privacy and regulatory protections attached to emotionally sensitive disclosures."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "Future work should test similar safety cues in functioning chatbot interfaces and examine behavioral outcomes such as click-through to help resources or actual referral use."
Zhang, Z., Lu, X., Zhang, Y. et al. (2026). Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1934264 - the article this story is about — the whole article — the passage: "A responsible chatbot should clearly state at onboarding that it cannot diagnose mental disorders, provide psychotherapy, replace counselors or physicians, or handle emergencies. Boundary messages should also appear again when users mention severe distress, self-harm, harm to others, medical decisions, or prolonged functional impairment."

Do not take this as professional medical advice.

The findings of other studies mentioned here are known to us through this document, which is the one we read; we did not open each of those studies.