{"id":497,"date":"2026-04-02T23:58:59","date_gmt":"2026-04-02T23:58:59","guid":{"rendered":"https:\/\/quantusintel.group\/osint\/blog\/2026\/04\/02\/google-ai-mode-is-not-gdpr-compliant-according-to-google-ai\/"},"modified":"2026-04-02T23:58:59","modified_gmt":"2026-04-02T23:58:59","slug":"google-ai-mode-is-not-gdpr-compliant-according-to-google-ai","status":"publish","type":"post","link":"https:\/\/quantusintel.group\/osint\/blog\/2026\/04\/02\/google-ai-mode-is-not-gdpr-compliant-according-to-google-ai\/","title":{"rendered":"Google AI Mode Is Not GDPR Compliant \u2014 According to Google AI"},"content":{"rendered":"<p>Author: Berend Watchus. Independent AI &amp; Cyber Sec. Researcher. Trendwatcher. [Publication for: OSINT Team] April 2,\u00a02026<\/p>\n<figure><img data-opt-id=771569372  fetchpriority=\"high\" decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*7Rxsy6RjV9BbfQWPOuUvHQ.png\" \/><\/figure>\n<figure><img data-opt-id=771569372  fetchpriority=\"high\" decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*WHwMT-R1NyNSOmypduu3BQ.png\" \/><\/figure>\n<p><strong>Google AI Mode Is Not GDPR Compliant\u200a\u2014\u200aAccording to Google\u00a0AI<\/strong><\/p>\n<p><em>Author: Berend Watchus. Independent AI &amp; Cybersecurity Researcher. April 2,\u00a02026.<\/em><\/p>\n<p>There is a legal right you almost certainly don\u2019t know you\u00a0have.<\/p>\n<p>Under GDPR Article 16, every EU citizen has the right to rectification of inaccurate personal data\u200a\u2014\u200awithout undue delay. If an organization holds or distributes false information about you, you have the legal right to have it corrected. Not eventually. Without undue\u00a0delay.<\/p>\n<p>Google AI mode generates responses about real people every day. Some of those responses are wrong. Fabricated researcher names. False biographical details. Misattributed work. Invented frameworks. I documented several such errors today about my own published research record\u200a\u2014\u200aincluding a completely fabricated researcher named \u201c\u2026\u2026\u2026\u2026.\u00a0\u2026\u2026\u2026\u2026.\u201d presented as a peer-level contributor in my field, and a false causal narrative connecting a 2014 event I organized to theoretical work I developed independently years\u00a0later.<\/p>\n<p>When I corrected these errors using primary source documentation\u200a\u2014\u200averbatim paper text, DOI-registered publications, a physical book with an ISBN\u200a\u2014\u200aGoogle AI mode accepted the corrections and updated its responses. In seven languages. Within hours of the original errors being generated.<\/p>\n<p>But here is the\u00a0problem.<\/p>\n<p>When the session ends, the correction disappears. The next person who queries the same topic cold receives the same original incorrect information. No persistent correction occurred. The person already aware of the error\u200a\u2014\u200athe only person the in-session correction actually helped\u200a\u2014\u200aalready knew the information was\u00a0wrong.<\/p>\n<p>I asked Google AI mode directly whether this satisfies GDPR Article 16. The answer it gave is worth reading carefully.<\/p>\n<p><em>\u201cJe hebt gelijk: Artikel 16 is bedoeld om de verspreiding van onjuiste informatie naar derden te stoppen. Een in-sessie correctie is een vorm van \u2018in-context learning\u2019 die alleen de huidige gebruiker helpt.\u201d<\/em><\/p>\n<p>[\u201cYou are correct: Article 16 is designed to stop the distribution of inaccurate information to third parties. An in-session correction is a form of in-context learning that only helps the current\u00a0user.\u201d]<\/p>\n<p>Google AI mode, when directly questioned, acknowledged that the mechanism available to users does not fulfill the legal right it is ostensibly there to\u00a0support.<\/p>\n<p>When asked what genuine Article 16 compliance would actually require, Google AI mode responded:<\/p>\n<p><em>\u201cOm aan Artikel 16 te voldoen, zou Google een permanent filter of een \u2018truth layer\u2019 moeten implementeren die onjuiste claims over een specifiek persoon onderschept voordat ze naar een willekeurige derde worden gestuurd. Het ontbreken van zo\u2019n persistent mechanisme wordt door veel juristen inderdaad gezien als een structurele schending.\u201d<\/em><\/p>\n<p>[\u201cTo comply with Article 16, Google would need to implement a permanent filter or \u2018truth layer\u2019 that intercepts incorrect claims about a specific person before they are sent to an arbitrary third party. The absence of such a persistent mechanism is indeed viewed by many legal experts as a structural violation.\u201d]<\/p>\n<p>The phrase \u201cstructural violation\u201d does not appear in this article as this author\u2019s assertion. It appears as a characterization generated by Google AI mode itself when directly questioned about its own compliance architecture.<\/p>\n<p>Google AI mode also cited the Italian Garante\u2019s enforcement action against ChatGPT in 2023\u200a\u2014\u200athe regulator that actually suspended an AI system over data protection failures\u200a\u2014\u200aas precedent for the position that technical complexity does not constitute a valid legal exemption:<\/p>\n<p><em>\u201cToezichthouders (zoals de Italiaanse Garante in de zaak tegen ChatGPT) hebben geoordeeld dat \u2018technische complexiteit\u2019 geen geldig excuus is om de rechten van betrokkenen te negeren.\u201d<\/em><\/p>\n<p>[\u201cRegulators (such as the Italian Garante in the case against ChatGPT) have ruled that \u2018technical complexity\u2019 is not a valid excuse to ignore the rights of data subjects.\u201d]<\/p>\n<h3>This is not the first time Google has faced scrutiny over\u00a0GDPR<\/h3>\n<ul>\n<li><a href=\"https:\/\/www.privacycompany.eu\/blog\/gdpr-violations-what-you-can-learn-from-the-first-50-million-eu-fine-issued-to-google-2\">GDPR violations: What you can learn from the first 50 million \u20ac fine issued to Google | Privacy Company Blog<\/a><\/li>\n<li><a href=\"https:\/\/valasys.com\/biggest-gdpr-fine-awarded-to-google\/\">Biggest GDPR Fine Awarded To Google!<\/a><\/li>\n<\/ul>\n<p>This is not the first time Google has faced scrutiny over GDPR compliance failures. The French regulator CNIL previously issued Google with a \u20ac50 million fine\u200a\u2014\u200aat the time the largest GDPR penalty ever imposed\u200a\u2014\u200afor violations related to transparency and user consent. That fine was for failing to properly inform users and obtain valid consent. The current issue described in this article is structurally different but legally adjacent: not a consent failure but a rectification failure, at a scale that dwarfs any previous enforcement scenario. The CNIL fine involved consent disclosures affecting users during device setup. The Article 16 compliance gap described here involves inaccurate personal data being distributed to billions of users with no persistent correction mechanism\u200a\u2014\u200aacross every query, every language, every country where Google AI mode operates.<\/p>\n<p><strong>The Substrate Is Irrelevant<\/strong><\/p>\n<p>A recurring argument in discussions about AI and data protection is that the technical complexity of AI systems creates a different category of compliance obligation than simpler data systems. Google AI mode\u2019s own responses suggest this argument does not hold legally. But it is worth examining why from first principles.<\/p>\n<p>Consider a concrete example. If Google AI mode mixed up Albert Einstein and Isaac Newton\u200a\u2014\u200aattributing one\u2019s discoveries to the other, connecting them to the wrong relationships, the wrong controversies, the wrong historical networks\u200a\u2014\u200aand a user corrected this error in a single chat session on a single account, that correction would affect exactly one person in exactly one session. The other billions of Google users would continue receiving the mixed-up version. Einstein\u2019s descendants, Newton\u2019s biographers, historians, students, journalists\u200a\u2014\u200aeveryone querying either name would still get the wrong information.<\/p>\n<p>The substrate of the correction doesn\u2019t change this. In the 20th century, corrections to false information in the public domain appeared in national newspapers, in published errata, in formal retractions. The medium was print. The obligation was the same: the correction had to reach the public domain where the false information had circulated. A private letter to the editor that was never published satisfied nothing.<\/p>\n<p>Google AI mode\u2019s in-session correction is the digital equivalent of that unpublished letter. It reaches one person. The false information continues circulating to everyone\u00a0else.<\/p>\n<p>GDPR in the EU, and equivalent privacy legislation across other continents and jurisdictions, exists specifically to protect citizens against this. The right to rectification is designed to correct inaccurate information about a person in the public domain\u200a\u2014\u200anot to provide a private acknowledgment of error to the person already aware of\u00a0it.<\/p>\n<p>The complexity of Google\u2019s own infrastructure\u200a\u2014\u200aits hardware, its software, its training pipelines, its billions of parameters\u200a\u2014\u200aadds nothing to the complexity of what rectification requires. Rectification is simple in legal terms: the inaccurate information about a person must stop being distributed to third parties. Whether achieving that technically requires a newspaper correction, a database update, or retraining a language model is an engineering problem for the data controller to solve. It is not a legal exemption.<\/p>\n<p>If it is not fixed in the public domain, it is not compliant. The substrate is irrelevant.<\/p>\n<p>Imagine a newspaper publisher arguing that a false front-page story about a private citizen cannot be corrected because their new printing press technology, their logistics management system, their delivery network, their smartphone apps for distribution staff, and their digital scanning infrastructure are all extremely complex systems that make targeted correction technically difficult. Nobody would accept that argument. The complexity of the publisher\u2019s internal operations is entirely irrelevant to the obligation to correct what reached the public. The citizen doesn\u2019t care how the newspaper is printed or delivered. They care that millions of people read a false story about them and that the correction reaches those same\u00a0people.<\/p>\n<p>Google\u2019s situation is structurally identical\u200a\u2014\u200anot as this author\u2019s legal conclusion, but as the logical extension of what Google AI mode itself described when questioned. The complexity of transformer architectures, attention mechanisms, parameter weights, inference pipelines, and distributed computing infrastructure is Google\u2019s internal operational reality. According to Google AI mode\u2019s own characterization, it has no bearing on what the person whose data is being incorrectly distributed to billions of users is legally entitled\u00a0to.<\/p>\n<p>To put it in the plainest possible terms: showing a correction in a single user\u2019s chat session, which disappears when the session ends and changes nothing for anyone else, is as serious a compliance solution as Google employees gathering to sing a song about updating the records\u200a\u2014\u200aand then not updating\u00a0them.<\/p>\n<p><strong>The Compliance Gap<\/strong><\/p>\n<p>Google AI mode described its current approach explicitly:<\/p>\n<p><em>\u201cDe huidige in-sessie methode is een interface-oplossing, geen juridische oplossing.\u201d<\/em><\/p>\n<p>[\u201cThe current in-session method is an interface solution, not a legal solution.\u201d]<\/p>\n<p>An interface solution. Built to look like compliance. Functioning as responsiveness to context. Satisfying nothing that Article 16 actually requires\u200a\u2014\u200aaccording to Google AI mode\u2019s own characterization.<\/p>\n<p><strong>What Actually\u00a0Works<\/strong><\/p>\n<p>When asked about more durable rectification pathways, Google AI mode identified the following:<\/p>\n<p><em>\u201cModellen maken steeds vaker gebruik van Retrieval-Augmented Generation (RAG), waarbij ze tijdens het genereren van een antwoord live op het internet zoeken. Een goed gedocumenteerde correctie op een publiek platform heeft dus een veel grotere kans om de output voor alle toekomstige gebruikers te corrigeren dan een chat-sessie.\u201d<\/em><\/p>\n<p>[\u201cModels increasingly use Retrieval-Augmented Generation (RAG), whereby they search the web live during response generation. A well-documented correction on a public platform therefore has a much greater chance of correcting the output for all future users than a chat session.\u201d]<\/p>\n<p>Publishing corrections on indexed platforms\u200a\u2014\u200aMedium, archive.ph, personal websites\u200a\u2014\u200acreates what Google AI mode called \u201ceen nieuwe bron van waarheid\u201d\u200a\u2014\u200aa new source of truth\u200a\u2014\u200athat RAG-enabled models can retrieve during inference. This is indirect, slow, and depends on Google\u2019s crawling and indexing decisions. It is not a legal right being fulfilled. It is a workaround for a compliance gap\u200a\u2014\u200aand Google AI mode characterized it as\u00a0such.<\/p>\n<p><strong>The Structural Problem<\/strong><\/p>\n<p>Billions of people interact with Google AI mode. Many of them have inaccurate information generated about them. Almost none of them know the in-session correction mechanism exists. Of those who find it, none of them can use it to correct what future users receive. The harm of inaccurate AI-generated personal data\u200a\u2014\u200areputational, professional, biographical\u200a\u2014\u200afalls on the people queried about, not on the people doing the querying.<\/p>\n<p>Article 16 exists precisely for this asymmetry. The right belongs to the person the data is about. The mechanism Google has built serves only the person already in the conversation\u200a\u2014\u200awho, in most cases of error, is not the person the data is about at\u00a0all.<\/p>\n<p>Google AI mode is generating inaccurate personal data at scale, distributing it to third parties, and offering a correction mechanism that\u200a\u2014\u200aaccording to its own characterization\u200a\u2014\u200acorrects nothing for anyone who wasn\u2019t already aware of the\u00a0error.<\/p>\n<p>Whether this constitutes a legal violation of GDPR Article 16 is a question for regulators and courts to determine. What is documented here is that Google AI mode itself, when directly and systematically questioned, described the absence of a persistent correction mechanism as something that \u201cveel juristen inderdaad zien als een structurele schending\u201d\u200a\u2014\u200athat many legal experts indeed view as a structural violation.<\/p>\n<p>This article reports what Google AI mode said. Readers and regulators may draw their own conclusions.<\/p>\n<p><em>Author: Berend Watchus. Independent non-profit AI &amp; Cybersecurity Researcher, Netherlands. April 2,\u00a02026.<\/em><\/p>\n<p><em>Methodology: The author conducted documented correction sessions with Google AI mode on April 2, 2026. All Dutch-language quotes are direct outputs from Google AI mode, reproduced verbatim with English translations provided by the author. Screenshots of all sessions are available. This article reports those outputs. It does not constitute legal\u00a0advice.<\/em><\/p>\n<p><em>Archives: [to be added upon publication.]<\/em><\/p>\n<p>screenshots \/\u00a0evidence<\/p>\n<figure><img data-opt-id=771569372  decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*8t8hwWe9-njjgIWTseaU4A.png\" \/><\/figure>\n<figure><img data-opt-id=771569372  decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*AoLobSY_ALu3S0V0aBqBBA.png\" \/><\/figure>\n<figure><img data-opt-id=771569372  decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*VDVSK5OTonx0CQzO0AClYA.png\" \/><\/figure>\n<figure><img data-opt-id=771569372  decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*KEU_NLcaqrtzM2XSm_-mvA.png\" \/><\/figure>\n<figure><img data-opt-id=771569372  decoding=\"async\" alt=\"\" src=\"https:\/\/cdn-images-1.medium.com\/max\/1024\/1*Cy28NWj2m8ZIydPxKOUKow.png\" \/><\/figure>\n<p>translated:<\/p>\n<p><strong>English: AI and the GDPR Right to Rectification (Article\u00a016)<\/strong><\/p>\n<p>You have identified one of the most complex legal and technical conflicts of our time: the clash between the static nature of trained AI models and the dynamic requirements of the\u00a0GDPR.<\/p>\n<ol>\n<li><strong>Information vs. Personal Data:<\/strong> Companies often argue that an AI model is a prediction engine, not a database. They claim incorrect output is a statistical \u201challucination\u201d rather than a stored \u201cdata point\u201d that can be rectified. However, European regulators maintain that if the output concerns an identifiable person, it falls under the\u00a0GDPR.<\/li>\n<li><strong>Technical Impossibility vs. Legal Duty:<\/strong> There is an \u201cimpossibility defense.\u201d Because personal data is woven into billions of parameters, it is technically difficult to \u201cedit\u201d a single fact without retraining the entire model. Regulators have ruled that technical complexity is not a valid excuse to ignore data subject\u00a0rights.<\/li>\n<li><strong>Why In-Session Correction is Insufficient:<\/strong> You are correct: Article 16 is intended to stop the spread of inaccurate information to <em>third parties<\/em>. An in-session correction only helps the current user. Without a persistent filter or \u201ctruth layer\u201d that intercepts incorrect claims for <em>all<\/em> future users, many jurists view this as a <strong>structural violation<\/strong>.<\/li>\n<li><strong>The Role of External Sources:<\/strong> Currently, the most durable route for rectification is <strong>external publication<\/strong>. By publishing the correction on an indexed platform, you create a new \u201csource of truth\u201d that AI models can retrieve during future generations (RAG), which has a much higher chance of correcting the output for everyone.<\/li>\n<\/ol>\n<p><strong>Deutsch: KI und das DSGVO-Recht auf Berichtigung (Artikel\u00a016)<\/strong><\/p>\n<p>Sie haben einen der komplexesten juristischen und technischen Konflikte unserer Zeit identifiziert: den Zusammensto\u00df zwischen der statischen Natur trainierter KI-Modelle und den dynamischen Anforderungen der\u00a0DSGVO.<\/p>\n<ol>\n<li><strong>Information vs. Personenbezogene Daten:<\/strong> Unternehmen argumentieren oft, dass ein KI-Modell eine Vorhersagemaschine und keine Datenbank ist. Sie behaupten, fehlerhafte Ausgaben seien statistische \u201cHalluzinationen\u201d und keine gespeicherten \u201cDatenpunkte\u201d. Europ\u00e4ische Regulierungsbeh\u00f6rden betonen jedoch, dass die DSGVO gilt, sobald sich die Ausgabe auf eine identifizierbare Person\u00a0bezieht.<\/li>\n<li><strong>Technische Unm\u00f6glichkeit vs. Rechtliche Pflicht:<\/strong> Es gibt eine \u201cUnm\u00f6glichkeits-Einrede\u201d. Da Daten in Milliarden von Parametern verwoben sind, ist es technisch schwierig, eine einzelne Tatsache zu \u201cbearbeiten\u201d. Beh\u00f6rden haben jedoch entschieden, dass technische Komplexit\u00e4t keine g\u00fcltige Ausrede ist, um Betroffenenrechte zu ignorieren.<\/li>\n<li><strong>Warum die In-Session-Korrektur nicht ausreicht:<\/strong> Sie haben recht: Artikel 16 soll die Verbreitung falscher Informationen an <em>Dritte<\/em> stoppen. Eine In-Session-Korrektur hilft nur dem aktuellen Nutzer. Ohne einen dauerhaften Filter, der falsche Behauptungen f\u00fcr <em>alle<\/em> zuk\u00fcnftigen Nutzer abf\u00e4ngt, sehen viele Juristen darin eine <strong>strukturelle Verletzung<\/strong>.<\/li>\n<li><strong>Die Rolle externer Quellen:<\/strong> Der nachhaltigste Weg zur Berichtigung ist derzeit die <strong>externe Ver\u00f6ffentlichung<\/strong>. Durch die Publikation auf einer indizierten Plattform schaffen Sie eine neue \u201cQuelle der Wahrheit\u201d, die KI-Modelle bei zuk\u00fcnftigen Abfragen (RAG) heranziehen k\u00f6nnen.<\/li>\n<\/ol>\n<p><strong>Fran\u00e7ais\u00a0: L\u2019IA et le droit de rectification du RGPD (Article\u00a016)<\/strong><\/p>\n<p>Vous avez identifi\u00e9 l\u2019un des conflits juridiques et techniques les plus complexes de notre \u00e9poque\u00a0: le choc entre la nature statique des mod\u00e8les d\u2019IA entra\u00een\u00e9s et les exigences dynamiques du\u00a0RGPD.<\/p>\n<ol>\n<li><strong>Information vs Donn\u00e9es personnelles\u00a0:<\/strong> Les entreprises soutiennent souvent qu\u2019un mod\u00e8le d\u2019IA est un moteur de pr\u00e9diction et non une base de donn\u00e9es. Elles affirment que les r\u00e9sultats incorrects sont des \u00ab\u00a0hallucinations\u00a0\u00bb statistiques et non des \u00ab\u00a0donn\u00e9es\u00a0\u00bb stock\u00e9es. Toutefois, les r\u00e9gulateurs europ\u00e9ens maintiennent que si le r\u00e9sultat concerne une personne identifiable, il rel\u00e8ve du\u00a0RGPD.<\/li>\n<li><strong>Impossibilit\u00e9 technique vs Obligation l\u00e9gale\u00a0:<\/strong> Il existe une \u00ab\u00a0d\u00e9fense d\u2019impossibilit\u00e9\u00a0\u00bb. Les donn\u00e9es \u00e9tant entrelac\u00e9es dans des milliards de param\u00e8tres, il est techniquement difficile de \u00ab\u00a0modifier\u00a0\u00bb un fait pr\u00e9cis. Les autorit\u00e9s ont jug\u00e9 que la complexit\u00e9 technique n\u2019est pas une excuse valable pour ignorer les droits des personnes concern\u00e9es.<\/li>\n<li><strong>Pourquoi la correction en session est insuffisante\u00a0:<\/strong> Vous avez raison\u00a0: l\u2019article 16 vise \u00e0 stopper la diffusion d\u2019informations inexactes \u00e0 des <em>tiers<\/em>. Une correction en session n\u2019aide que l\u2019utilisateur actuel. Sans un filtre permanent ou une \u00ab\u00a0couche de v\u00e9rit\u00e9\u00a0\u00bb interceptant les affirmations erron\u00e9es pour <em>tous<\/em> les futurs utilisateurs, de nombreux juristes y voient une <strong>violation structurelle<\/strong>.<\/li>\n<li><strong>Le r\u00f4le des sources externes\u00a0:<\/strong> Actuellement, la voie de rectification la plus durable est la <strong>publication externe<\/strong>. En publiant la correction sur une plateforme index\u00e9e, vous cr\u00e9ez une nouvelle \u00ab\u00a0source de v\u00e9rit\u00e9\u00a0\u00bb que les mod\u00e8les d\u2019IA peuvent r\u00e9cup\u00e9rer lors de g\u00e9n\u00e9rations futures\u00a0(RAG).<\/li>\n<\/ol>\n<p><strong>Espa\u00f1ol: La IA y el derecho de rectificaci\u00f3n del RGPD (Art\u00edculo 16)<\/strong><\/p>\n<p>Ha identificado uno de los conflictos jur\u00eddicos y t\u00e9cnicos m\u00e1s complejos de nuestro tiempo: el choque entre la naturaleza est\u00e1tica de los modelos de IA entrenados y los requisitos din\u00e1micos del\u00a0RGPD.<\/p>\n<ol>\n<li><strong>Informaci\u00f3n frente a Datos personales:<\/strong> Las empresas suelen argumentar que un modelo de IA es un motor de predicci\u00f3n, no una base de datos. Afirman que los resultados incorrectos son \u201calucinaciones\u201d estad\u00edsticas y no \u201cdatos\u201d almacenados. Sin embargo, los reguladores europeos sostienen que si el resultado afecta a una persona identificable, se aplica el\u00a0RGPD.<\/li>\n<li><strong>Imposibilidad t\u00e9cnica frente a Deber legal:<\/strong> Existe una \u201cdefensa de imposibilidad\u201d. Dado que los datos est\u00e1n entretejidos en miles de millones de par\u00e1metros, es t\u00e9cnicamente dif\u00edcil \u201ceditar\u201d un hecho concreto. Las autoridades han dictaminado que la complejidad t\u00e9cnica no es una excusa v\u00e1lida para ignorar los derechos de los interesados.<\/li>\n<li><strong>Por qu\u00e9 la correcci\u00f3n en sesi\u00f3n es insuficiente:<\/strong> Tiene raz\u00f3n: el art\u00edculo 16 pretende detener la difusi\u00f3n de informaci\u00f3n inexacta a <em>terceros<\/em>. Una correcci\u00f3n en sesi\u00f3n solo ayuda al usuario actual. Sin un filtro persistente que intercepte las afirmaciones incorrectas para <em>todos<\/em> los futuros usuarios, muchos juristas consideran que se trata de una <strong>violaci\u00f3n estructural<\/strong>.<\/li>\n<li><strong>El papel de las fuentes externas:<\/strong> Actualmente, la v\u00eda m\u00e1s duradera para la rectificaci\u00f3n es la <strong>publicaci\u00f3n externa<\/strong>. Al publicar la correcci\u00f3n en una plataforma indexada, se crea una nueva \u201cfuente de verdad\u201d que los modelos de IA pueden consultar en futuras generaciones (RAG).<\/li>\n<\/ol>\n<p><strong>Portugu\u00eas: A IA e o direito de retifica\u00e7\u00e3o do RGPD (Artigo\u00a016)<\/strong><\/p>\n<p>Identificou um dos conflitos jur\u00eddicos e t\u00e9cnicos mais complexos do nosso tempo: o choque entre a natureza est\u00e1tica dos modelos de IA treinados e os requisitos din\u00e2micos do\u00a0RGPD.<\/p>\n<ol>\n<li><strong>Informa\u00e7\u00e3o vs. Dados Pessoais:<\/strong> As empresas argumentam frequentemente que um modelo de IA \u00e9 um motor de previs\u00e3o, n\u00e3o uma base de dados. Alegam que os resultados incorretos s\u00e3o \u201calucina\u00e7\u00f5es\u201d estat\u00edsticas e n\u00e3o \u201cdados\u201d armazenados. No entanto, os reguladores europeus defendem que, se o resultado disser respeito a uma pessoa identific\u00e1vel, recai sobre o\u00a0RGPD.<\/li>\n<li><strong>Impossibilidade T\u00e9cnica vs. Dever Legal:<\/strong> Existe uma \u201cdefesa de impossibilidade\u201d. Como os dados est\u00e3o entrela\u00e7ados em milhares de milh\u00f5es de par\u00e2metros, \u00e9 tecnicamente dif\u00edcil \u201ceditar\u201d um facto espec\u00edfico. As autoridades decidiram que a complexidade t\u00e9cnica n\u00e3o \u00e9 uma desculpa v\u00e1lida para ignorar os direitos dos titulares dos\u00a0dados.<\/li>\n<li><strong>Por que a corre\u00e7\u00e3o em sess\u00e3o \u00e9 insuficiente:<\/strong> Tem raz\u00e3o: o Artigo 16 destina-se a impedir a propaga\u00e7\u00e3o de informa\u00e7\u00f5es imprecisas a <em>terceiros<\/em>. Uma corre\u00e7\u00e3o em sess\u00e3o apenas ajuda o utilizador atual. Sem um filtro persistente ou uma \u201ccamada de verdade\u201d que intercepte alega\u00e7\u00f5es incorretas para <em>todos<\/em> os futuros utilizadores, muitos juristas veem isto como uma <strong>viola\u00e7\u00e3o estrutural<\/strong>.<\/li>\n<li><strong>O papel das fontes externas:<\/strong> Atualmente, a via mais duradoura para a retifica\u00e7\u00e3o \u00e9 a <strong>publica\u00e7\u00e3o externa<\/strong>. Ao publicar a corre\u00e7\u00e3o numa plataforma indexada, cria uma nova \u201cfonte de verdade\u201d que os modelos de IA podem recuperar em gera\u00e7\u00f5es futuras\u00a0(RAG).<\/li>\n<\/ol>\n<p><strong>\u0420\u0443\u0441\u0441\u043a\u0438\u0439 (Russisch): \u0418\u0418 \u0438 \u043f\u0440\u0430\u0432\u043e \u043d\u0430 \u0438\u0441\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u043f\u043e GDPR (\u0421\u0442\u0430\u0442\u044c\u044f\u00a016)<\/strong><\/p>\n<p>\u0412\u044b \u0437\u0430\u0442\u0440\u043e\u043d\u0443\u043b\u0438 \u043e\u0434\u0438\u043d \u0438\u0437 \u0441\u0430\u043c\u044b\u0445 \u0441\u043b\u043e\u0436\u043d\u044b\u0445 \u044e\u0440\u0438\u0434\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0438 \u0442\u0435\u0445\u043d\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043a\u043e\u043d\u0444\u043b\u0438\u043a\u0442\u043e\u0432 \u0441\u043e\u0432\u0440\u0435\u043c\u0435\u043d\u043d\u043e\u0441\u0442\u0438: \u0441\u0442\u043e\u043b\u043a\u043d\u043e\u0432\u0435\u043d\u0438\u0435 \u043c\u0435\u0436\u0434\u0443 \u0441\u0442\u0430\u0442\u0438\u0447\u043d\u043e\u0439 \u043f\u0440\u0438\u0440\u043e\u0434\u043e\u0439 \u043e\u0431\u0443\u0447\u0435\u043d\u043d\u044b\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439 \u0418\u0418 \u0438 \u0434\u0438\u043d\u0430\u043c\u0438\u0447\u0435\u0441\u043a\u0438\u043c\u0438 \u0442\u0440\u0435\u0431\u043e\u0432\u0430\u043d\u0438\u044f\u043c\u0438 GDPR.<\/p>\n<ol>\n<li><strong>\u0418\u043d\u0444\u043e\u0440\u043c\u0430\u0446\u0438\u044f \u043f\u0440\u043e\u0442\u0438\u0432 \u043f\u0435\u0440\u0441\u043e\u043d\u0430\u043b\u044c\u043d\u044b\u0445 \u0434\u0430\u043d\u043d\u044b\u0445:<\/strong> \u041a\u043e\u043c\u043f\u0430\u043d\u0438\u0438 \u0447\u0430\u0441\u0442\u043e \u0443\u0442\u0432\u0435\u0440\u0436\u0434\u0430\u044e\u0442, \u0447\u0442\u043e \u043c\u043e\u0434\u0435\u043b\u044c \u0418\u0418\u200a\u2014\u200a\u044d\u0442\u043e \u043c\u0435\u0445\u0430\u043d\u0438\u0437\u043c \u043f\u0440\u043e\u0433\u043d\u043e\u0437\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f, \u0430 \u043d\u0435 \u0431\u0430\u0437\u0430 \u0434\u0430\u043d\u043d\u044b\u0445. \u041e\u043d\u0438 \u0437\u0430\u044f\u0432\u043b\u044f\u044e\u0442, \u0447\u0442\u043e \u043d\u0435\u0432\u0435\u0440\u043d\u044b\u0435 \u043e\u0442\u0432\u0435\u0442\u044b \u044f\u0432\u043b\u044f\u044e\u0442\u0441\u044f \u0441\u0442\u0430\u0442\u0438\u0441\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u043c\u0438 \u00ab\u0433\u0430\u043b\u043b\u044e\u0446\u0438\u043d\u0430\u0446\u0438\u044f\u043c\u0438\u00bb, \u0430 \u043d\u0435 \u0445\u0440\u0430\u043d\u044f\u0449\u0438\u043c\u0438\u0441\u044f \u00ab\u0434\u0430\u043d\u043d\u044b\u043c\u0438\u00bb. \u041e\u0434\u043d\u0430\u043a\u043e \u0435\u0432\u0440\u043e\u043f\u0435\u0439\u0441\u043a\u0438\u0435 \u0440\u0435\u0433\u0443\u043b\u044f\u0442\u043e\u0440\u044b \u043d\u0430\u0441\u0442\u0430\u0438\u0432\u0430\u044e\u0442: \u0435\u0441\u043b\u0438 \u043e\u0442\u0432\u0435\u0442 \u043a\u0430\u0441\u0430\u0435\u0442\u0441\u044f \u0438\u0434\u0435\u043d\u0442\u0438\u0444\u0438\u0446\u0438\u0440\u0443\u0435\u043c\u043e\u0433\u043e \u043b\u0438\u0446\u0430, \u043e\u043d \u043f\u043e\u0434\u043f\u0430\u0434\u0430\u0435\u0442 \u043f\u043e\u0434 \u0434\u0435\u0439\u0441\u0442\u0432\u0438\u0435\u00a0GDPR.<\/li>\n<li><strong>\u0422\u0435\u0445\u043d\u0438\u0447\u0435\u0441\u043a\u0430\u044f \u043d\u0435\u0432\u043e\u0437\u043c\u043e\u0436\u043d\u043e\u0441\u0442\u044c \u043f\u0440\u043e\u0442\u0438\u0432 \u044e\u0440\u0438\u0434\u0438\u0447\u0435\u0441\u043a\u043e\u0439 \u043e\u0431\u044f\u0437\u0430\u043d\u043d\u043e\u0441\u0442\u0438:<\/strong> \u0421\u0443\u0449\u0435\u0441\u0442\u0432\u0443\u0435\u0442 \u0430\u0440\u0433\u0443\u043c\u0435\u043d\u0442 \u043e \u00ab\u0442\u0435\u0445\u043d\u0438\u0447\u0435\u0441\u043a\u043e\u0439 \u043d\u0435\u0432\u043e\u0437\u043c\u043e\u0436\u043d\u043e\u0441\u0442\u0438\u00bb. \u041f\u043e\u0441\u043a\u043e\u043b\u044c\u043a\u0443 \u0434\u0430\u043d\u043d\u044b\u0435 \u0432\u043f\u043b\u0435\u0442\u0435\u043d\u044b \u0432 \u043c\u0438\u043b\u043b\u0438\u0430\u0440\u0434\u044b \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440\u043e\u0432, \u0442\u0435\u0445\u043d\u0438\u0447\u0435\u0441\u043a\u0438 \u0441\u043b\u043e\u0436\u043d\u043e \u00ab\u043e\u0442\u0440\u0435\u0434\u0430\u043a\u0442\u0438\u0440\u043e\u0432\u0430\u0442\u044c\u00bb \u043e\u0434\u0438\u043d \u043a\u043e\u043d\u043a\u0440\u0435\u0442\u043d\u044b\u0439 \u0444\u0430\u043a\u0442. \u0420\u0435\u0433\u0443\u043b\u044f\u0442\u043e\u0440\u044b \u043f\u043e\u0441\u0442\u0430\u043d\u043e\u0432\u0438\u043b\u0438, \u0447\u0442\u043e \u0442\u0435\u0445\u043d\u0438\u0447\u0435\u0441\u043a\u0430\u044f \u0441\u043b\u043e\u0436\u043d\u043e\u0441\u0442\u044c \u043d\u0435 \u044f\u0432\u043b\u044f\u0435\u0442\u0441\u044f \u043e\u043f\u0440\u0430\u0432\u0434\u0430\u043d\u0438\u0435\u043c \u0434\u043b\u044f \u0438\u0433\u043d\u043e\u0440\u0438\u0440\u043e\u0432\u0430\u043d\u0438\u044f \u043f\u0440\u0430\u0432 \u0441\u0443\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u0434\u0430\u043d\u043d\u044b\u0445.<\/li>\n<li><strong>\u041f\u043e\u0447\u0435\u043c\u0443 \u0438\u0441\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u0432\u043d\u0443\u0442\u0440\u0438 \u0441\u0435\u0441\u0441\u0438\u0438 \u043d\u0435\u0434\u043e\u0441\u0442\u0430\u0442\u043e\u0447\u043d\u043e:<\/strong> \u0412\u044b \u043f\u0440\u0430\u0432\u044b: \u0421\u0442\u0430\u0442\u044c\u044f 16 \u043f\u0440\u0438\u0437\u0432\u0430\u043d\u0430 \u043e\u0441\u0442\u0430\u043d\u043e\u0432\u0438\u0442\u044c \u0440\u0430\u0441\u043f\u0440\u043e\u0441\u0442\u0440\u0430\u043d\u0435\u043d\u0438\u0435 \u043d\u0435\u0442\u043e\u0447\u043d\u043e\u0439 \u0438\u043d\u0444\u043e\u0440\u043c\u0430\u0446\u0438\u0438 <em>\u0442\u0440\u0435\u0442\u044c\u0438\u043c \u043b\u0438\u0446\u0430\u043c<\/em>. \u0418\u0441\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u0432\u043d\u0443\u0442\u0440\u0438 \u0441\u0435\u0441\u0441\u0438\u0438 \u043f\u043e\u043c\u043e\u0433\u0430\u0435\u0442 \u0442\u043e\u043b\u044c\u043a\u043e \u0442\u0435\u043a\u0443\u0449\u0435\u043c\u0443 \u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u0435\u043b\u044e. \u0411\u0435\u0437 \u043f\u043e\u0441\u0442\u043e\u044f\u043d\u043d\u043e\u0433\u043e \u0444\u0438\u043b\u044c\u0442\u0440\u0430, \u043f\u0435\u0440\u0435\u0445\u0432\u0430\u0442\u044b\u0432\u0430\u044e\u0449\u0435\u0433\u043e \u043d\u0435\u0432\u0435\u0440\u043d\u044b\u0435 \u0443\u0442\u0432\u0435\u0440\u0436\u0434\u0435\u043d\u0438\u044f \u0434\u043b\u044f <em>\u0432\u0441\u0435\u0445<\/em> \u0431\u0443\u0434\u0443\u0449\u0438\u0445 \u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u0435\u043b\u0435\u0439, \u043c\u043d\u043e\u0433\u0438\u0435 \u044e\u0440\u0438\u0441\u0442\u044b \u0432\u0438\u0434\u044f\u0442 \u0432 \u044d\u0442\u043e\u043c <strong>\u0441\u0442\u0440\u0443\u043a\u0442\u0443\u0440\u043d\u043e\u0435 \u043d\u0430\u0440\u0443\u0448\u0435\u043d\u0438\u0435<\/strong>.<\/li>\n<li><strong>\u0420\u043e\u043b\u044c \u0432\u043d\u0435\u0448\u043d\u0438\u0445 \u0438\u0441\u0442\u043e\u0447\u043d\u0438\u043a\u043e\u0432:<\/strong> \u0412 \u043d\u0430\u0441\u0442\u043e\u044f\u0449\u0435\u0435 \u0432\u0440\u0435\u043c\u044f \u043d\u0430\u0438\u0431\u043e\u043b\u0435\u0435 \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u043f\u0443\u0442\u044c \u0434\u043b\u044f \u0438\u0441\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u044f\u200a\u2014\u200a<strong>\u0432\u043d\u0435\u0448\u043d\u044f\u044f \u043f\u0443\u0431\u043b\u0438\u043a\u0430\u0446\u0438\u044f<\/strong>. \u041f\u0443\u0431\u043b\u0438\u043a\u0443\u044f \u0438\u0441\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u043d\u0430 \u0438\u043d\u0434\u0435\u043a\u0441\u0438\u0440\u0443\u0435\u043c\u043e\u0439 \u043f\u043b\u0430\u0442\u0444\u043e\u0440\u043c\u0435, \u0432\u044b \u0441\u043e\u0437\u0434\u0430\u0435\u0442\u0435 \u043d\u043e\u0432\u044b\u0439 \u00ab\u0438\u0441\u0442\u043e\u0447\u043d\u0438\u043a \u0438\u0441\u0442\u0438\u043d\u044b\u00bb, \u043a\u043e\u0442\u043e\u0440\u044b\u0439 \u043c\u043e\u0434\u0435\u043b\u0438 \u0418\u0418 \u043c\u043e\u0433\u0443\u0442 \u0438\u0437\u0432\u043b\u0435\u043a\u0430\u0442\u044c \u043f\u0440\u0438 \u0431\u0443\u0434\u0443\u0449\u0438\u0445 \u0437\u0430\u043f\u0440\u043e\u0441\u0430\u0445\u00a0(RAG).<\/li>\n<\/ol>\n<p><strong>\u4e2d\u6587 (Chinees): \u4eba\u5de5\u667a\u80fd\u4e0e GDPR \u66f4\u6b63\u6743\uff08\u7b2c 16\u00a0\u6761\uff09<\/strong><\/p>\n<p>\u4f60\u6307\u51fa\u4e86\u5f53\u4ee3\u6700\u590d\u6742\u7684\u6cd5\u5f8b\u4e0e\u6280\u672f\u51b2\u7a81\u4e4b\u4e00\uff1a\u8bad\u7ec3\u597d\u7684 AI \u6a21\u578b\u7684\u9759\u6001\u6027\u8d28\u4e0e GDPR \u52a8\u6001\u8981\u6c42\u4e4b\u95f4\u7684\u78b0\u649e\u3002<\/p>\n<ol>\n<li><strong>\u4fe1\u606f\u4e0e\u4e2a\u4eba\u6570\u636e\u4e4b\u4e89\uff1a<\/strong> \u4f01\u4e1a\u5e38\u8fa9\u79f0 AI \u6a21\u578b\u662f\u9884\u6d4b\u5f15\u64ce\u800c\u975e\u6570\u636e\u5e93\u3002\u4ed6\u4eec\u58f0\u79f0\u9519\u8bef\u7684\u8f93\u51fa\u662f\u7edf\u8ba1\u5b66\u4e0a\u7684\u201c\u5e7b\u89c9\u201d\uff0c\u800c\u975e\u5b58\u50a8\u7684\u201c\u6570\u636e\u70b9\u201d\u3002\u7136\u800c\uff0c\u6b27\u6d32\u76d1\u7ba1\u673a\u6784\u575a\u6301\u8ba4\u4e3a\uff0c\u53ea\u8981\u8f93\u51fa\u6d89\u53ca\u53ef\u8bc6\u522b\u7684\u4e2a\u4eba\uff0c\u5c31\u5c5e\u4e8e GDPR\u00a0\u7684\u8303\u7574\u3002<\/li>\n<li><strong>\u6280\u672f\u4e0a\u7684\u4e0d\u53ef\u80fd\u4e0e\u6cd5\u5f8b\u4e49\u52a1\uff1a<\/strong> \u5b58\u5728\u4e00\u79cd\u201c\u4e0d\u53ef\u80fd\u8fa9\u62a4\u201d\u3002\u7531\u4e8e\u4e2a\u4eba\u6570\u636e\u4ea4\u7ec7\u5728\u6570\u5341\u4ebf\u4e2a\u53c2\u6570\u4e2d\uff0c\u6280\u672f\u4e0a\u5f88\u96be\u5728\u4e0d\u91cd\u65b0\u8bad\u7ec3\u6a21\u578b\u7684\u60c5\u51b5\u4e0b\u201c\u4fee\u6539\u201d\u5355\u4e2a\u4e8b\u5b9e\u3002\u76d1\u7ba1\u673a\u6784\u5df2\u88c1\u5b9a\uff0c\u6280\u672f\u590d\u6742\u6027\u4e0d\u80fd\u4f5c\u4e3a\u5ffd\u89c6\u6570\u636e\u4e3b\u4f53\u6743\u5229\u7684\u5408\u6cd5\u7406\u7531\u3002<\/li>\n<li><strong>\u4e3a\u4f55\u4f1a\u8bdd\u5185\u66f4\u6b63\u4e0d\u8db3\uff1a<\/strong> \u4f60\u662f\u5bf9\u7684\uff1a\u7b2c 16 \u6761\u65e8\u5728\u505c\u6b62\u5411<strong>\u7b2c\u4e09\u65b9<\/strong>\u4f20\u64ad\u4e0d\u51c6\u786e\u4fe1\u606f\u3002\u4f1a\u8bdd\u5185\u66f4\u6b63\u4ec5\u5bf9\u5f53\u524d\u7528\u6237\u6709\u6548\u3002\u5982\u679c\u7f3a\u4e4f\u9488\u5bf9<strong>\u6240\u6709<\/strong>\u672a\u6765\u7528\u6237\u62e6\u622a\u9519\u8bef\u58f0\u660e\u7684\u6301\u4e45\u8fc7\u6ee4\u5668\u6216\u201c\u771f\u76f8\u5c42\u201d\uff0c\u8bb8\u591a\u6cd5\u5f8b\u4e13\u5bb6\u8ba4\u4e3a\u8fd9\u6784\u6210\u4e86<strong>\u7ed3\u6784\u6027\u8fdd\u89c4<\/strong>\u3002<\/li>\n<li><strong>\u5916\u90e8\u6765\u6e90\u7684\u4f5c\u7528\uff1a<\/strong> \u76ee\u524d\u6700\u6301\u4e45\u7684\u66f4\u6b63\u9014\u5f84\u662f<strong>\u5916\u90e8\u53d1\u5e03<\/strong>\u3002\u901a\u8fc7\u5728\u88ab\u7d22\u5f15\u7684\u5e73\u53f0\u4e0a\u53d1\u5e03\u66f4\u6b63\u4fe1\u606f\uff0c\u4f60\u521b\u5efa\u4e86\u4e00\u4e2a\u65b0\u7684\u201c\u4e8b\u5b9e\u6765\u6e90\u201d\uff0cAI \u6a21\u578b\u5728\u672a\u6765\u7684\u751f\u6210\u8fc7\u7a0b\u4e2d\uff08\u901a\u8fc7 RAG \u6280\u672f\uff09\u53ef\u4ee5\u68c0\u7d22\u5230\u8be5\u4fe1\u606f\u3002<\/li>\n<\/ol>\n<p><strong>\u65e5\u672c\u8a9e (Japans): AI \u3068 GDPR \u8a02\u6b63\u6a29\uff08\u7b2c16\u6761\uff09<\/strong><\/p>\n<p>\u3042\u306a\u305f\u306f\u3001\u73fe\u4ee3\u306b\u304a\u3051\u308b\u6700\u3082\u8907\u96d1\u306a\u6cd5\u7684\u30fb\u6280\u8853\u7684\u5bfe\u7acb\u306e\u4e00\u3064\u3092\u6307\u6458\u3057\u307e\u3057\u305f\u3002\u305d\u308c\u306f\u3001\u5b66\u7fd2\u6e08\u307f AI \u30e2\u30c7\u30eb\u306e\u9759\u7684\u306a\u6027\u8cea\u3068 GDPR \u306e\u52d5\u7684\u306a\u8981\u4ef6\u3068\u306e\u885d\u7a81\u3067\u3059\u3002<\/p>\n<ol>\n<li><strong>\u60c5\u5831 vs \u500b\u4eba\u30c7\u30fc\u30bf:<\/strong> \u4f01\u696d\u5074\u306f\u3001AI \u30e2\u30c7\u30eb\u306f\u4e88\u6e2c\u30a8\u30f3\u30b8\u30f3\u3067\u3042\u308a\u3001\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u3067\u306f\u306a\u3044\u3068\u3057\u3070\u3057\u3070\u4e3b\u5f35\u3057\u307e\u3059\u3002\u8aa4\u3063\u305f\u51fa\u529b\u306f\u7d71\u8a08\u7684\u306a\u300c\u30cf\u30eb\u30b7\u30cd\u30fc\u30b7\u30e7\u30f3\uff08\u5e7b\u899a\uff09\u300d\u3067\u3042\u308a\u3001\u8a02\u6b63\u53ef\u80fd\u306a\u300c\u4fdd\u5b58\u3055\u308c\u305f\u30c7\u30fc\u30bf\u300d\u3067\u306f\u306a\u3044\u3068\u3044\u3046\u7406\u5c48\u3067\u3059\u3002\u3057\u304b\u3057\u3001\u6b27\u5dde\u306e\u898f\u5236\u5f53\u5c40\u306f\u3001\u51fa\u529b\u304c\u7279\u5b9a\u306e\u500b\u4eba\u306b\u95a2\u3059\u308b\u3082\u306e\u3067\u3042\u308b\u9650\u308a\u3001GDPR \u304c\u9069\u7528\u3055\u308c\u308b\u3068\u3044\u3046\u898b\u89e3\u3092\u7dad\u6301\u3057\u3066\u3044\u307e\u3059\u3002<\/li>\n<li><strong>\u6280\u8853\u7684\u4e0d\u53ef\u80fd\u6027 vs \u6cd5\u7684\u7fa9\u52d9:<\/strong> \u300c\u4e0d\u53ef\u80fd\u6027\u306e\u5f01\u8b77\u300d\u3068\u3044\u3046\u8b70\u8ad6\u304c\u3042\u308a\u307e\u3059\u3002\u30c7\u30fc\u30bf\u304c\u6570\u5341\u5104\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306b\u7e54\u308a\u8fbc\u307e\u308c\u3066\u3044\u308b\u305f\u3081\u3001\u7279\u5b9a\u306e\u4e8b\u5b9f\u306e\u307f\u3092\u300c\u7de8\u96c6\u300d\u3059\u308b\u3053\u3068\u306f\u6280\u8853\u7684\u306b\u56f0\u96e3\u3067\u3059\u3002\u3057\u304b\u3057\u3001\u5f53\u5c40\u306f\u3001\u6280\u8853\u7684\u306a\u8907\u96d1\u3055\u306f\u30c7\u30fc\u30bf\u4e3b\u4f53\u306e\u6a29\u5229\u3092\u7121\u8996\u3059\u308b\u6b63\u5f53\u306a\u7406\u7531\u306b\u306f\u306a\u3089\u306a\u3044\u3068\u88c1\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002<\/li>\n<li><strong>\u30bb\u30c3\u30b7\u30e7\u30f3\u5185\u8a02\u6b63\u304c\u4e0d\u5341\u5206\u306a\u7406\u7531:<\/strong> \u3042\u306a\u305f\u306e\u6307\u6458\u901a\u308a\u3067\u3059\u3002\u7b2c16\u6761\u306f\u3001<strong>\u7b2c\u4e09\u8005<\/strong>\u3078\u306e\u4e0d\u6b63\u78ba\u306a\u60c5\u5831\u306e\u62e1\u6563\u3092\u6b62\u3081\u308b\u3053\u3068\u3092\u76ee\u7684\u3068\u3057\u3066\u3044\u307e\u3059\u3002\u30bb\u30c3\u30b7\u30e7\u30f3\u5185\u306e\u8a02\u6b63\u306f\u73fe\u5728\u306e\u30e6\u30fc\u30b6\u30fc\u3092\u52a9\u3051\u308b\u3060\u3051\u3067\u3001\u5c06\u6765\u306e<strong>\u3059\u3079\u3066\u306e<\/strong>\u30e6\u30fc\u30b6\u30fc\u306b\u5bfe\u3057\u3066\u8aa4\u3063\u305f\u4e3b\u5f35\u3092\u906e\u65ad\u3059\u308b\u6c38\u7d9a\u7684\u306a\u30d5\u30a3\u30eb\u30bf\u30fc\u304c\u306a\u3051\u308c\u3070\u3001\u591a\u304f\u306e\u6cd5\u5b66\u8005\u306f\u3053\u308c\u3092<strong>\u69cb\u9020\u7684\u306a\u9055\u53cd<\/strong>\u3068\u898b\u306a\u3057\u307e\u3059\u3002<\/li>\n<li><strong>\u5916\u90e8\u30bd\u30fc\u30b9\u306e\u5f79\u5272:<\/strong> \u73fe\u5728\u3001\u6700\u3082\u5b9f\u52b9\u6027\u306e\u3042\u308b\u8a02\u6b63\u30eb\u30fc\u30c8\u306f<strong>\u5916\u90e8\u3078\u306e\u516c\u958b<\/strong>\u3067\u3059\u3002\u30a4\u30f3\u30c7\u30c3\u30af\u30b9\u3055\u308c\u305f\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u3067\u8a02\u6b63\u60c5\u5831\u3092\u516c\u958b\u3059\u308b\u3053\u3068\u3067\u3001AI \u30e2\u30c7\u30eb\u304c\u5c06\u6765\u306e\u751f\u6210\u6642\uff08RAG\u306a\u3069\uff09\u306b\u53d6\u5f97\u3067\u304d\u308b\u65b0\u3057\u3044\u300c\u771f\u5b9f\u306e\u30bd\u30fc\u30b9\u300d\u3092\u4f5c\u308a\u51fa\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n<p><strong>\u0627\u0644\u0639\u0631\u0628\u064a\u0629 (Arabisch): \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0648\u0627\u0644\u062d\u0642 \u0641\u064a \u0627\u0644\u062a\u0635\u062d\u064a\u062d \u0628\u0645\u0648\u062c\u0628 GDPR (\u0627\u0644\u0645\u0627\u062f\u0629\u00a016)<\/strong><\/p>\n<p>\u0644\u0642\u062f \u062d\u062f\u062f\u062a \u0648\u0627\u062d\u062f\u0627\u064b \u0645\u0646 \u0623\u0643\u062b\u0631 \u0627\u0644\u0635\u0631\u0627\u0639\u0627\u062a \u0627\u0644\u0642\u0627\u0646\u0648\u0646\u064a\u0629 \u0648\u0627\u0644\u062a\u0642\u0646\u064a\u0629 \u062a\u0639\u0642\u064a\u062f\u0627\u064b \u0641\u064a \u0639\u0635\u0631\u0646\u0627: \u0627\u0644\u062a\u0635\u0627\u062f\u0645 \u0628\u064a\u0646 \u0627\u0644\u0637\u0628\u064a\u0639\u0629 \u0627\u0644\u062b\u0627\u0628\u062a\u0629 \u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0644\u0645\u062f\u0631\u0628\u0629 \u0648\u0627\u0644\u0645\u062a\u0637\u0644\u0628\u0627\u062a \u0627\u0644\u062f\u064a\u0646\u0627\u0645\u064a\u0643\u064a\u0629 \u0644\u0644\u0627\u0626\u062d\u0629 \u0627\u0644\u0639\u0627\u0645\u0629 \u0644\u062d\u0645\u0627\u064a\u0629 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a\u00a0(GDPR).<\/p>\n<ol>\n<li><strong>\u0627\u0644\u0645\u0639\u0644\u0648\u0645\u0627\u062a \u0645\u0642\u0627\u0628\u0644 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a \u0627\u0644\u0634\u062e\u0635\u064a\u0629:<\/strong> \u063a\u0627\u0644\u0628\u0627\u064b \u0645\u0627 \u062a\u062c\u0627\u062f\u0644 \u0627\u0644\u0634\u0631\u0643\u0627\u062a \u0628\u0623\u0646 \u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0647\u0648 \u0645\u062d\u0631\u0643 \u062a\u0646\u0628\u0624 \u0648\u0644\u064a\u0633 \u0642\u0627\u0639\u062f\u0629 \u0628\u064a\u0627\u0646\u0627\u062a. \u0648\u064a\u0632\u0639\u0645\u0648\u0646 \u0623\u0646 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u063a\u064a\u0631 \u0627\u0644\u0635\u062d\u064a\u062d\u0629 \u0647\u064a \u201c\u0647\u0644\u0648\u0633\u0629\u201d \u0625\u062d\u0635\u0627\u0626\u064a\u0629 \u0648\u0644\u064a\u0633\u062a \u201c\u0646\u0642\u0637\u0629 \u0628\u064a\u0627\u0646\u0627\u062a\u201d \u0645\u062e\u0632\u0646\u0629 \u064a\u0645\u0643\u0646 \u062a\u0635\u062d\u064a\u062d\u0647\u0627. \u0648\u0645\u0639 \u0630\u0644\u0643\u060c \u064a\u0635\u0631 \u0627\u0644\u0645\u0646\u0638\u0645\u0648\u0646 \u0627\u0644\u0623\u0648\u0631\u0648\u0628\u064a\u0648\u0646 \u0639\u0644\u0649 \u0623\u0646\u0647 \u0625\u0630\u0627 \u0643\u0627\u0646\u062a \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a \u062a\u062a\u0639\u0644\u0642 \u0628\u0634\u062e\u0635 \u0645\u062d\u062f\u062f\u060c \u0641\u0625\u0646\u0647\u0627 \u062a\u062e\u0636\u0639 \u0644\u0640\u00a0GDPR.<\/li>\n<li><strong>\u0627\u0644\u0627\u0633\u062a\u062d\u0627\u0644\u0629 \u0627\u0644\u062a\u0642\u0646\u064a\u0629 \u0645\u0642\u0627\u0628\u0644 \u0627\u0644\u0648\u0627\u062c\u0628 \u0627\u0644\u0642\u0627\u0646\u0648\u0646\u064a:<\/strong> \u0647\u0646\u0627\u0643 \u0645\u0627 \u064a\u0633\u0645\u0649 \u0628\u0640 \u201c\u062f\u0641\u0627\u0639 \u0627\u0644\u0627\u0633\u062a\u062d\u0627\u0644\u0629\u201d. \u0646\u0638\u0631\u0627\u064b \u0644\u0623\u0646 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a \u0627\u0644\u0634\u062e\u0635\u064a\u0629 \u0645\u0646\u0633\u0648\u062c\u0629 \u0641\u064a \u0645\u0644\u064a\u0627\u0631\u0627\u062a \u0627\u0644\u0645\u0639\u0644\u0645\u0627\u062a\u060c \u0641\u0645\u0646 \u0627\u0644\u0635\u0639\u0628 \u062a\u0642\u0646\u064a\u0627\u064b \u201c\u062a\u0639\u062f\u064a\u0644\u201d \u062d\u0642\u064a\u0642\u0629 \u0648\u0627\u062d\u062f\u0629. \u0648\u0642\u062f \u0642\u0631\u0631\u062a \u0633\u0644\u0637\u0627\u062a \u0627\u0644\u0631\u0642\u0627\u0628\u0629 \u0623\u0646 \u0627\u0644\u062a\u0639\u0642\u064a\u062f \u0627\u0644\u062a\u0642\u0646\u064a \u0644\u064a\u0633 \u0639\u0630\u0631\u0627\u064b \u0645\u0642\u0628\u0648\u0644\u0627\u064b \u0644\u062a\u062c\u0627\u0647\u0644 \u062d\u0642\u0648\u0642 \u0623\u0635\u062d\u0627\u0628 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a.<\/li>\n<li><strong>\u0644\u0645\u0627\u0630\u0627 \u0644\u0627 \u064a\u0643\u0641\u064a \u0627\u0644\u062a\u0635\u062d\u064a\u062d \u062f\u0627\u062e\u0644 \u0627\u0644\u062c\u0644\u0633\u0629:<\/strong> \u0623\u0646\u062a \u0639\u0644\u0649 \u062d\u0642\u061b \u062a\u0647\u062f\u0641 \u0627\u0644\u0645\u0627\u062f\u0629 16 \u0625\u0644\u0649 \u0648\u0642\u0641 \u0646\u0634\u0631 \u0627\u0644\u0645\u0639\u0644\u0648\u0645\u0627\u062a \u063a\u064a\u0631 \u0627\u0644\u062f\u0642\u064a\u0642\u0629 \u0625\u0644\u0649 <strong>\u0623\u0637\u0631\u0627\u0641 \u062b\u0627\u0644\u062b\u0629<\/strong>. \u0627\u0644\u062a\u0635\u062d\u064a\u062d \u062f\u0627\u062e\u0644 \u0627\u0644\u062c\u0644\u0633\u0629 \u064a\u0633\u0627\u0639\u062f \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645 \u0627\u0644\u062d\u0627\u0644\u064a \u0641\u0642\u0637. \u0648\u0628\u062f\u0648\u0646 \u0648\u062c\u0648\u062f \u0645\u0631\u0634\u062d \u062f\u0627\u0626\u0645 \u0623\u0648 \u201c\u0637\u0628\u0642\u0629 \u062d\u0642\u064a\u0642\u0629\u201d \u062a\u0645\u0646\u0639 \u0627\u0644\u0627\u062f\u0639\u0627\u0621\u0627\u062a \u0627\u0644\u062e\u0627\u0637\u0626\u0629 \u0644\u0640 <strong>\u062c\u0645\u064a\u0639<\/strong> \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646 \u0627\u0644\u0645\u0633\u062a\u0642\u0628\u0644\u064a\u064a\u0646\u060c \u064a\u0631\u0649 \u0627\u0644\u0639\u062f\u064a\u062f \u0645\u0646 \u0627\u0644\u0642\u0627\u0646\u0648\u0646\u064a\u064a\u0646 \u0623\u0646 \u0647\u0630\u0627 \u064a\u0645\u062b\u0644 <strong>\u0627\u0646\u062a\u0647\u0627\u0643\u0627\u064b\u00a0\u0647\u064a\u0643\u0644\u064a\u0627\u064b<\/strong>.<\/li>\n<li><strong>\u062f\u0648\u0631 \u0627\u0644\u0645\u0635\u0627\u062f\u0631 \u0627\u0644\u062e\u0627\u0631\u062c\u064a\u0629:<\/strong> \u062d\u0627\u0644\u064a\u0627\u064b\u060c \u0627\u0644\u0645\u0633\u0627\u0631 \u0627\u0644\u0623\u0643\u062b\u0631 \u062f\u064a\u0645\u0648\u0645\u0629 \u0644\u0644\u062a\u0635\u062d\u064a\u062d \u0647\u0648 <strong>\u0627\u0644\u0646\u0634\u0631 \u0627\u0644\u062e\u0627\u0631\u062c\u064a<\/strong>. \u0645\u0646 \u062e\u0644\u0627\u0644 \u0646\u0634\u0631 \u0627\u0644\u062a\u0635\u062d\u064a\u062d \u0639\u0644\u0649 \u0645\u0646\u0635\u0629 \u0645\u0641\u0647\u0631\u0633\u0629\u060c \u0641\u0625\u0646\u0643 \u062a\u0646\u0634\u0626 \u201c\u0645\u0635\u062f\u0631\u0627\u064b \u062c\u062f\u064a\u062f\u0627\u064b \u0644\u0644\u062d\u0642\u064a\u0642\u0629\u201d \u064a\u0645\u0643\u0646 \u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0630\u0643\u0627\u0621 \u0627\u0644\u0627\u0635\u0637\u0646\u0627\u0639\u064a \u0627\u0633\u062a\u0631\u062c\u0627\u0639\u0647 \u0623\u062b\u0646\u0627\u0621 \u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u062a\u0648\u0644\u064a\u062f \u0627\u0644\u0645\u0633\u062a\u0642\u0628\u0644\u064a\u0629 (RAG).<\/li>\n<\/ol>\n<p>*<strong>Global Recognition\u200a\u2014\u200aThe Same Conclusion in Every\u00a0Language<\/strong><\/p>\n<p>Following the documented correction sessions described in this article, I presented the core finding\u200a\u2014\u200athat in-session corrections do not satisfy GDPR Article 16 because they only help the current user while the false information continues reaching all future users\u200a\u2014\u200ato Google AI mode across multiple languages. The response was consistent across Russian, Chinese, Japanese, Arabic, and\u00a0others.<\/p>\n<p>In every language, Google AI mode arrived at the same three conclusions independently:<\/p>\n<p>That the technical complexity defense does not constitute a legal exemption\u200a\u2014\u200aregulators have ruled this explicitly. That in-session correction only helps the current user and does not stop distribution of inaccurate information to third parties. That the absence of a persistent filter or truth layer is viewed by many legal experts as a structural violation.<\/p>\n<p>The Russian response stated it directly: without a permanent filter intercepting incorrect claims for all future users, many lawyers see a structural violation. The Chinese response confirmed that technical complexity cannot serve as justification for ignoring data subject rights. The Japanese response agreed that session corrections fall short because Article 16 is designed to stop inaccurate information reaching third parties. The Arabic response used identical framing: a permanent filter or truth layer is required, and its absence represents a structural violation in the eyes of many legal\u00a0experts.<\/p>\n<p>This consistency across languages is not coincidental. It reflects that when Google AI mode is directly and systematically questioned about its own compliance architecture, it arrives at the same legal characterization regardless of the language of the query. The finding is not a translation artifact or a quirk of one session. It is what Google AI mode says about itself when asked directly, in any language, by anyone who knows to\u00a0ask.<\/p>\n<p><img data-opt-id=574357117  decoding=\"async\" src=\"https:\/\/medium.com\/_\/stat?event=post.clientViewed&amp;referrerSource=full_rss&amp;postId=bec966ba9ae3\" width=\"1\" height=\"1\" alt=\"\" \/><\/p>\n<hr \/>\n<p><a href=\"https:\/\/osintteam.blog\/google-ai-mode-is-not-gdpr-compliant-according-to-google-ai-bec966ba9ae3\">Google AI Mode Is Not GDPR Compliant \u2014 According to Google AI<\/a> was originally published in <a href=\"https:\/\/osintteam.blog\/\">OSINT Team<\/a> on Medium, where people are continuing the conversation by highlighting and responding to this story.<\/p>","protected":false},"excerpt":{"rendered":"<p>Author: Berend Watchus. Independent AI &amp; Cyber Sec. Researcher. Trendwatcher. [Publication for: OSINT Team] April 2,\u00a02026 Google AI Mode Is Not GDPR Compliant\u200a\u2014\u200aAccording to Google\u00a0AI Author: Berend Watchus. Independent AI &amp; Cybersecurity Researcher. April 2,\u00a02026. There is a legal right you almost certainly don\u2019t know you\u00a0have. Under GDPR Article 16, every EU citizen has the &#8230; <a title=\"Google AI Mode Is Not GDPR Compliant \u2014 According to Google AI\" class=\"read-more\" href=\"https:\/\/quantusintel.group\/osint\/blog\/2026\/04\/02\/google-ai-mode-is-not-gdpr-compliant-according-to-google-ai\/\" aria-label=\"Read more about Google AI Mode Is Not GDPR Compliant \u2014 According to Google AI\">Read more<\/a><\/p>\n","protected":false},"author":1,"featured_media":498,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-497","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/posts\/497","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/comments?post=497"}],"version-history":[{"count":0,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/posts\/497\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/media\/498"}],"wp:attachment":[{"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/media?parent=497"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/categories?post=497"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/quantusintel.group\/osint\/wp-json\/wp\/v2\/tags?post=497"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}