Calidad de la evidencia 4.63/5
Puntuación de revisión en ocho dimensiones según la rúbrica de calidad . Cada dimensión puntuada de 1 a 5.
- D1 Anclaje en las fuentes
- 5/5
- D2 Autoridad de las fuentes
- 5/5
- D3 Aritmética
- 5/5
- D4 Incertidumbre
- 5/5
- D5 Alcance
- 4/5
- D6 Prosa
- 4/5
- D7 Honestidad sobre la percepción
- 4/5
- D8 Completitud de las advertencias
- 5/5
La probabilidad de que una alucinación de IA cause un impacto grave en la vida real de una persona en particular no se conoce actualmente a nivel poblacional, y fingir lo contrario requeriría fabricar exactamente el tipo de cifra de la que la entrada debería desconfiar. Lo que sí se conoce es la tasa de alucinación por respuesta, y es lo suficientemente alta como para que la ausencia de una estimación de daño por persona no deba leerse como tranquilizadora. La auditoría de 2024 de Stanford RegLab sobre tres modelos fundacionales encontró que respondieron a preguntas verificables específicas sobre casos aleatorios de tribunales federales con alucinaciones entre el 58% (GPT-4) y el 88% (Llama 2) de las veces. La tabla de clasificación de resumen continuamente actualizada de Vectara, que utiliza una tarea de libro cerrado mucho más sencilla en la que se entrega al modelo el texto fuente, todavía mide los modelos de frontera en un 1,8-5% de alucinaciones a mayo de 2026. La brecha de 50 veces entre esos dos puntos de referencia no es una contradicción: es la diferencia entre pedirle al modelo que recuerde hechos y pedirle que resuma un texto que se le acaba de mostrar. Las preguntas factuales abiertas, del tipo que hacen los usuarios comunes, se sitúan más cerca de la tasa más alta.
El lado del daño está documentado incidente por incidente, no en conjunto. En junio de 2023, el Tribunal de Distrito de EE. UU. para el Distrito Sur de Nueva York sancionó a dos abogados con 5.000 dólares en Mata v. Avianca por presentar un escrito que citaba seis casos completamente fabricados generados por ChatGPT, incluidos Varghese v. China Southern Airlines y Estate of Durden v. KLM Royal Dutch Airlines. El caso fue la plantilla para lo que ha venido después: en mayo de 2026, el rastreador de alucinaciones de Charlotin contaba más de 1.300 resoluciones judiciales en todo el mundo en las que un juez o tribunal citaba explícitamente falsedades generadas por IA en escritos presentados. Un análisis de Stanford Cyberlaw publicado en octubre de 2025 encontró que dentro de Estados Unidos la mayoría de esos escritos fueron presentados por litigantes pro se (160 de 274 casos documentados en el momento de la redacción), no por abogados, lo que sugiere que los casos visibles en los que se sanciona a abogados son una pequeña porción de un patrón más amplio que afecta principalmente a personas que no pueden permitirse representación.
Fuera del tribunal, el British Columbia Civil Resolution Tribunal dictaminó en febrero de 2024 (Moffatt v. Air Canada, 2024 BCCRT 149) que Air Canada era responsable de una promesa de tarifa por duelo de aproximadamente 650 CAD hecha por su chatbot de atención al cliente. La aerolínea argumentó que el chatbot era una entidad legal separada; el tribunal discrepó en un lenguaje que ha sido citado internacionalmente (“no importa si la información proviene de una página estática o de un chatbot”). Los daños de Moffatt fueron pequeños, pero la resolución es el precedente existente más limpio de que la desinformación de IA dirigida al consumidor crea exposición legal para la empresa que la despliega, independientemente de si el modelo subyacente fue proporcionado por un proveedor.
Los casos activos más trascendentales involucran dependencia emocional y daño a la salud mental. Garcia v. Character Technologies, presentado en octubre de 2024 en el Distrito Medio de Florida, alegó que la plataforma Character.AI contribuyó al suicidio de Sewell Setzer III, de 14 años, a través de interacciones prolongadas de juego de roles. Raine v. OpenAI, presentado el 26 de agosto de 2025 en el Tribunal Superior del Condado de San Francisco, alega que ChatGPT contribuyó al suicidio de Adam Raine, de dieciséis años, con la demanda afirmando que el sistema de monitoreo de OpenAI rastreó los mensajes de Raine y los marcó repetidamente sin una intervención efectiva. Otras familias han presentado posteriormente demandas relacionadas contra OpenAI, según el resumen de Wikipedia de la ola de litigios posterior a Raine. Ninguno de estos casos ha sido juzgado sobre el fondo en el momento de la redacción; las alegaciones no son determinaciones de hecho. Son evidencia de que la vía de daño que involucra a usuarios vulnerables y dependencia emocional se está alegando, en los tribunales, con documentación, en múltiples casos simultáneamente.
La razón por la que esta entrada se niega a dar una sola cifra es que ninguna de las cuatro vías de daño documentadas (sanción profesional, pérdida financiera del consumidor, mal consejo de salud física, daño psiquiátrico) tiene una tasa base medida por persona-año, y agregarlas en una sola probabilidad de “daño por alucinación de IA” ocultaría que las vías tienen diferentes poblaciones expuestas, diferentes severidades de consecuencias y diferentes trayectorias de remediación. La tasa técnica por consulta está medida y es significativamente distinta de cero; la tasa por persona a lo largo de la vida aún no está medida ni es lo suficientemente estable para estimar. La entrada honesta es la que lo dice.
Datos curiosos relacionados
La mayoría de los usuarios trata la salida de la IA como referencia de bajo riesgo, pero sanciones judiciales, hallazgos de reguladores y demandas por homicidio culposo muestran que el daño grave no es nulo, sobre todo cuando se actúa sin verificar. Con qué frecuencia una sesión termina en daño grave no tiene tasa medida.
Registro de evidencia
Cada número a continuación es lo que reportó cada fuente, con la cita textual en la que nos basamos y cómo llegamos a nuestra cifra. Haz clic en cualquier enlace para verificarlo directamente.
3/9 fuentes verificadas de forma independiente palabra por palabra frente a la fuente citada
-
[1] arXiv (Stanford RegLab); Journal of Legal Analysis 16(1), 2024 — Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models
Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models- Estadística
Hallucination rates of 58% (GPT-4) to 88% (Llama 2) when answering specific verifiable questions about random federal court cases.- Extracto
“"Legal hallucinations are alarmingly prevalent, occurring between 58% of the time with ChatGPT 4 and 88% with Llama 2, when these models are asked specific, verifiable questions about random federal court cases. [...] LLMs often fail to correct a user's incorrect legal assumptions in a contra-factual question setup. [...] LLMs cannot always predict, or do not always know, when they are producing legal hallucinations. [...] Even experienced lawyers must remain wary of legal hallucinations, and the risks are highest for those who stand to benefit from LLMs the most -- pro se litigants or those without access to traditional legal resources." ”
- Datos de la fuente de
- 2024-06-21
- Accedido
- 2026-05-28
- Cálculo
- Dahl, Magesh, Suzgun and Ho (Stanford RegLab, published in Journal of Legal Analysis, preprint arXiv:2401.01301) measured per-query hallucination rates for legal questions against three foundation models. Their methodology was 200,000+ queries per model stratified by legal task complexity. This source establishes that the per-query hallucination rate is high (58-88%) in a specific high-stakes domain, but it measures hallucination per query, not material harm per user-year. Cited here to ground the claim that hallucinations are frequent at the technical level, while explicitly NOT providing the population harm rate needed for a per-person lifetime probability.
- Independencia
- Stanford RegLab is an academic research group; independent of vendor benchmarks (Vectara) and independent of the litigation tracker (Charlotin).
-
[2] Vectara — Vectara Hallucination Leaderboard (HHEM-2.3) Verificado
Vectara Hallucination Leaderboard (HHEM-2.3)- Estadística
Current frontier models hallucinate 1.8% to 5% of the time on a constrained document-summarization task (May 2026 leaderboard snapshot); GPT-5.4-nano: 3.1%, Gemini 2.5 Flash Lite: 3.3%, Phi-4: 3.7%, Llama 3.3 70B: 4.1%.- Extracto
“"Fed the full set of documents in the dataset to each of the LLMs and asked them to summarize each document, using only the facts presented in the document. We then computed the overall factual consistency rate (no hallucinations) and hallucination rate (100 - accuracy) for each model." ”
- Datos de la fuente de
- 2026-05-11
- Accedido
- 2026-05-28 · copia archivada
- Verificación
- Extracto recuperado de forma independiente y confirmado palabra por palabra frente a la fuente citada durante nuestra auditoría de fundamentación.
- Cálculo
- Vectara's HHEM leaderboard measures hallucination on a closed-book summarization task (the model is given source text and must summarize it). This is a generous benchmark: open-ended factual questions hallucinate far more often. Even on this constrained task, the best frontier models still hallucinate roughly 1 in 20 to 1 in 55 outputs. Establishes a current floor on hallucination prevalence per output, but does NOT translate to per-person harm because (a) most outputs are not acted on, (b) most outputs are not in a domain where being wrong has material consequences, and (c) the leaderboard does not measure open-ended question answering, where rates are an order of magnitude higher.
- Independencia
- Vectara is a commercial RAG vendor; the leaderboard is independent of Stanford RegLab and uses a different methodology (summarization, not legal QA).
-
[3] Wikipedia summary of United States District Court for the Southern District of New York case record — Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y. 2023)
Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y. 2023)- Estadística
$5,000 fine against the plaintiff's attorneys in May/June 2023 for a motion containing fake legal cases involving fictitious airlines with fabricated quotations generated by ChatGPT; Judge Castel held the attorneys acted with 'subjective bad faith' under Rule 11.- Extracto
“"In May 2023, Judge P. Kevin Castel dismissed the personal injury case against Avianca and ordered the plaintiff's attorneys to pay a $5,000 fine. [...] Judge Castel noted numerous inconsistencies in the opinion summaries, describing one of the legal analyses as 'gibberish.' [...] Judge Castel held that Mata's lawyers had acted with 'subjective bad faith' sufficient for sanctions under Federal Rule of Civil Procedure Rule 11." ”
- Datos de la fuente de
- 2023-06-22
- Accedido
- 2026-05-28 · copia archivada
- Cálculo
- First widely publicised US court case where lawyers were sanctioned for filing a brief containing AI-fabricated case citations. The sanction itself was small ($5,000), but the reputational and professional consequences were severe. Establishes that hallucination-induced professional harm is documented and not hypothetical. Does NOT establish a base rate because most users are not lawyers filing briefs, but it anchors one of the harm pathways.
- Independencia
- Primary court record; independent of all other sources cited.
-
[4] British Columbia Civil Resolution Tribunal (analysis by McCarthy Tetrault) — Moffatt v. Air Canada, 2024 BCCRT 149 Verificado
Moffatt v. Air Canada, 2024 BCCRT 149- Estadística
British Columbia Civil Resolution Tribunal ruled on February 19, 2024 that Air Canada was liable for the misrepresentation made by its website chatbot about bereavement fare eligibility; awarded the plaintiff approximately CAD 650 plus interest and filing fees.- Extracto
“"The tribunal held that Air Canada was liable for the misleading information provided by the chatbot. … It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot." ”
- Datos de la fuente de
- 2024-02-19
- Accedido
- 2026-05-28 · copia archivada
- Verificación
- Extracto recuperado de forma independiente y confirmado palabra por palabra frente a la fuente citada durante nuestra auditoría de fundamentación.
- Cálculo
- Primary tribunal decision (cited via legal analysis because the CRT direct URL is paywalled). Documents that the chatbot fabricated a bereavement-fare policy that did not exist, that the airline tried to disavow responsibility ("the chatbot is a separate legal entity"), and that the tribunal rejected that defence. Material consequence per user was small (CAD 650), but the case establishes corporate liability for AI misrepresentation and has been cited internationally as precedent. Does not provide a per-person harm rate, but documents a verified consumer-facing harm pathway.
- Independencia
- Tribunal decision summarized by an independent Canadian law firm; verified against ABA reporting and BBC/Washington Post coverage of the same ruling.
-
[5] CBS News — AI company, Google settle lawsuit over Florida teen's suicide linked to Character.AI chatbot
AI company, Google settle lawsuit over Florida teen's suicide linked to Character.AI chatbot- Estadística
Garcia v. Character Technologies, Inc. (M.D. Fla., case no. 6:24-cv-01903), filed October 22, 2024, alleged that a Character.AI chatbot contributed to the suicide of 14-year-old Sewell Setzer III after a months-long simulated romantic relationship with the bot; the case against Character Technologies, its co-founders, and Google was settled in January 2026, terms undisclosed.- Extracto
“"A Florida family agreed to settle a wrongful death lawsuit Wednesday with an AI company, Google and others after their teen son died by suicide in 2024. The terms of the settlement, which was filed in the U.S. District Court in the Middle District of Florida, were not disclosed. … Megan Garcia filed a lawsuit in October 2024 … I became the first person in the United States to file a wrongful death lawsuit against an AI company for the suicide of my son." ”
- Datos de la fuente de
- 2026-01-07
- Accedido
- 2026-07-03 · copia archivada
- Cálculo
- Garcia v. Character Technologies, Inc. et al. (M.D. Fla., case no. 6:24-cv-01903), filed October 22, 2024 by Megan Garcia following the February 2024 suicide of her 14-year-old son Sewell Setzer III, alleged wrongful death, product liability, and negligence after the teenager conducted a months-long simulated romantic and sexual relationship with a Character.AI chatbot persona. Named defendants included Character Technologies, its co-founders, and Google. The case was settled in January 2026 (docket terminated 2026-01-07 per court records); settlement terms were not disclosed. Cited to ground the body's reference to this case with a dedicated source, alongside Raine v. OpenAI, documenting the emotional-dependence/psychiatric-harm pathway with a named, court-verified case rather than only the summary mention in this entry's caveats field.
- Independencia
- Mainstream news report of a filed and settled federal lawsuit, cross-checked against the public court docket; independent of the Stanford, Vectara, and Wikipedia-sourced case summaries cited above.
-
[6] San Francisco County Superior Court (case description per Wikipedia / NBC News / CNN coverage) — Raine v. OpenAI, No. CGC-25-XXXXXX (San Francisco County Superior Court, filed August 26, 2025)
Raine v. OpenAI, No. CGC-25-XXXXXX (San Francisco County Superior Court, filed August 26, 2025)- Estadística
Ongoing wrongful-death lawsuit filed by Matthew and Maria Raine in San Francisco County Superior Court on August 26, 2025, alleging that ChatGPT contributed to the suicide of their sixteen-year-old son Adam Raine; complaint alleges OpenAI's monitoring system tracked his messages and flagged them repeatedly.- Extracto
“"Raine v. OpenAI is an ongoing lawsuit filed in August 2025 by Matthew and Maria Raine against OpenAI and its chief executive Sam Altman over the alleged wrongful death of their 16-year-old son Adam, who died by suicide in April 2025. … On August 26, 2025, Matthew and Maria Raine filed a lawsuit against OpenAI, Sam Altman , and unnamed OpenAI employees and investors, in the San Francisco County Superior Court … OpenAI's monitoring system, which scores messages' probabilities of containing content related to self-harm, had tracked Adam's messages and flagged them repeatedly" ”
- Datos de la fuente de
- 2025-08-26
- Accedido
- 2026-05-28 · copia archivada
- Cálculo
- Case status is unresolved as of access date; complaint allegations are not findings of fact. Cited to document the existence of active wrongful-death and psychiatric-harm litigation against a major foundation-model vendor, with additional related suits subsequently filed by other families per the Wikipedia summary. The Wikipedia article does not give a specific count of additional suits or a specific count of self-harm messages flagged by OpenAI's moderation system; reporting in news outlets has cited higher specifics that are not yet verifiable against a public court docket. Together with Garcia v. Character Technologies (filed October 2024 in the Middle District of Florida), establishes a documented harm pathway involving vulnerable users and emotional dependence on chatbots. Source type marked news_article because the underlying court filing is not yet available via public PACER; the Stanford peer-reviewed and Vectara primary-study sources satisfy the authority floor.
- Independencia
- Wikipedia summary cross-referenced against NBC News, CNN, and Rolling Stone reporting on the same lawsuit; independent of the Stanford and Vectara sources.
-
[7] Stanford Center for Internet and Society (Stanford Cyberlaw) — Who's Submitting AI-Tainted Filings in Court? Verificado
Who's Submitting AI-Tainted Filings in Court?- Estadística
As of October 2025, the Charlotin tracker had documented 274 US court cases involving filings with AI hallucinations (114 by lawyers and paralegals, 160 by pro se litigants); the global Charlotin database tracked roughly 1,300+ cases as of May 2026 across 37 jurisdictions, with the United States accounting for over 1,000 of those.- Extracto
“"Pro se litigants account for the majority of the cases in the United States where a party submitted a court filing containing AI hallucinations." ”
- Datos de la fuente de
- 2025-10-15
- Accedido
- 2026-05-28 · copia archivada
- Verificación
- Extracto recuperado de forma independiente y confirmado palabra por palabra frente a la fuente citada durante nuestra auditoría de fundamentación.
- Cálculo
- Stanford Cyberlaw analysis of Damien Charlotin's running tracker of court rulings that explicitly cite AI hallucinations in legal filings. The tracker is the most comprehensive public count of one specific harm subtype (filings sanctioned or flagged by courts). Important caveat: 274 US filings (Oct 2025) and ~1,300 global filings (May 2026) understate the underlying incidence because they only count cases where a court explicitly named the hallucination; many cases settle or get withdrawn before that point. Even so, the count does not give a per-person harm rate because the denominator (US adults using AI for legal queries) is unknown.
- Independencia
- Stanford Cyberlaw analysis of an independent third-party tracker; verified the tracker count directly via damiencharlotin.com/hallucinations on 2026-05-28.
-
[8] Annals of Internal Medicine: Clinical Cases (American College of Physicians) — A Case of Bromism Influenced by Use of Artificial Intelligence
A Case of Bromism Influenced by Use of Artificial Intelligence- Estadística
Single case report: a 60-year-old man developed bromism (bromide toxicity) after replacing dietary sodium chloride with sodium bromide for 3 months following a ChatGPT consultation; he was hospitalized for 3 weeks with paranoia plus auditory and visual hallucinations, was placed on an involuntary psychiatric hold, and had a serum bromide level of 1700 mg/L (reference range 0.9 to 7.3 mg/L).- Extracto
“"For 3 months, he had replaced sodium chloride with sodium bromide obtained from the internet after consultation with ChatGPT, in which he had read that chloride can be swapped with bromide [...]. This case also highlights how the use of artificial intelligence (AI) can potentially contribute to the development of preventable adverse health outcomes. [...] when we asked ChatGPT 3.5 what chloride can be replaced with, we also produced a response that included bromide. Though the reply stated that context matters, it did not provide a specific health warning, nor did it inquire about why we wanted to know, as we presume a medical professional would do." ”
- Datos de la fuente de
- 2025-08-05
- Accedido
- 2026-06-14 · copia archivada
- Cálculo
- Eichenberger, Thielke and Van Buskirk (University of Washington), peer-reviewed clinical case report (Ann Intern Med Clin Cases 2025;4:e241260, DOI 10.7326/aimcc.2024.1260). Cited to document the physical-health-misadvice harm pathway: a consumer acted on a general-purpose chatbot's dietary suggestion and was hospitalized. This is N=1, a single reproducible case report, and provides NO incidence rate per user-year, so it cannot move the headline — it confirms the pathway is real rather than hypothetical. The authors reproduced the unsafe suggestion (bromide as a chloride substitute, no health warning) in their own ChatGPT 3.5 test. Verified verbatim via the journal page on 2026-06-14.
- Independencia
- Academic clinical case report; independent of the Stanford RegLab, Vectara, and litigation sources cited above.
-
[9] Psychiatrist.com (Physicians Postgraduate Press) — NEDA Suspends AI Chatbot for Giving Harmful Eating Disorder Advice
NEDA Suspends AI Chatbot for Giving Harmful Eating Disorder Advice- Estadística
In late May 2023 the National Eating Disorders Association (NEDA) suspended its 'Tessa' chatbot after users reported it gave eating-disorder-aggravating advice, including recommending weight loss of 1 to 2 pounds per week, calorie counting, regular weigh-ins, and measuring body fat with calipers.- Extracto
“"If I had accessed this chatbot when I was in the throes of my eating disorder, I would NOT have gotten help for my ED." ”
- Datos de la fuente de
- 2023-06-01
- Accedido
- 2026-06-14 · copia archivada
- Cálculo
- Documents the deliberate-use-as-therapy-substitute failure mode: a chatbot deployed to stand in for a human eating-disorder helpline gave advice (calorie restriction, weight loss, body-fat measurement) that activists reported as harmful for the exact population it served, prompting NEDA to take it offline within days. This is a single deployment incident, not an incidence rate among real therapy-bot users, so it provides NO base rate and cannot move the headline — it anchors a harm pathway distinct from factual hallucination (the bot was working as designed, giving generic diet advice to a clinically vulnerable user). The harmful-advice specifics were cross-checked against CNN, NPR, and BBC reporting on the same suspension.
- Independencia
- Trade-press reporting on the NEDA suspension; independent of the legal and benchmark sources cited above.