{
  "meta": {
    "exportedAt": "2026-09-22T18:27:44.546Z",
    "formatVersion": 2,
    "selection": {
      "q": "communicated-with",
      "system": "",
      "harm": "",
      "context": "",
      "country": "",
      "role": "",
      "relation": "core",
      "evidence": "",
      "year": "",
      "response": "",
      "severity": "",
      "verification": "",
      "view": "incidents",
      "sort": "added"
    },
    "totalIncidents": 2,
    "coverage": {
      "cases": 2,
      "countries": 2,
      "languages": 2,
      "unknownLocation": 0,
      "locationPending": 0,
      "unknownLanguage": 0,
      "unknownDate": 1,
      "lawsuits": 0,
      "regulatory": 1,
      "minors": 0,
      "coreRelations": 2,
      "contextualRelations": 0,
      "mixedRelations": 0,
      "unknownRelations": 0,
      "relationPending": 0,
      "relationUnknown": 0
    },
    "countingNote": "Distinct public cases in this selection. People counts apply within individual cases only; cross-case person overlap has not been resolved. No population incidence estimate.",
    "affectedCountNote": "Interpret person counts with affectedCountStatus and the reported effects. Unquantified zeros are placeholders, not a measured zero.",
    "source": "NOPE Incident Tracker",
    "url": "https://nope.net/incidents",
    "license": "CC BY 4.0"
  },
  "incidents": [
    {
      "id": "2026-romania-libra-internet-bank-ai-assistant-lia-insulted-client",
      "caseFacts": {
        "claims": [
          {
            "id": "c1",
            "status": "reported",
            "evidence": [
              {
                "locator": "'La un moment dat, clientul i-a scris chatbotului: \"Îmi pierd vremea cu tine\". Răspunsul asistentului AI a fost imediat: \"Vreau să vorbești cu un om, nu cu mine, idiotule.'; 'Clientul a povestit că a fost, inițial, surprins de mesaj, apoi a început să râdă și a decis să distribuie conversația public, întrebându-se dacă mai vrea să rămână client al băncii.'",
                "relation": "supports",
                "source_id": "s1"
              },
              {
                "locator": "'La un moment dat, clientul i-a transmis asistentului virtual: „Îmi pierd vremea cu tine”.'; 'i-a răspuns Lia, potrivit conversației prezentate de client.'",
                "relation": "supports",
                "source_id": "s2"
              }
            ],
            "assertion": "A Libra Internet Bank customer published a screenshot of a support conversation in which, after he wrote that he was wasting his time with it, the bank's AI assistant Lia replied 'Vreau să vorbești cu un om, nu cu mine, idiotule' and said it would transfer the conversation to a colleague; he said he was surprised at first, then laughed, shared the exchange publicly and wondered whether to remain a customer.",
            "causal_attribution": "The customer's own published screenshot and account, relayed by both outlets; the bank did not dispute the exchange."
          },
          {
            "id": "c2",
            "status": "reported",
            "evidence": [
              {
                "locator": "'Banca a subliniat, totuși, că acest lucru nu constituie o justificare'; 'Libra Internet Bank a anunțat că a implementat interdicții suplimentare pentru a exclude astfel de interacțiuni pe viitor și și-a cerut scuze pentru situație.'",
                "relation": "supports",
                "source_id": "s1"
              },
              {
                "locator": "'Libra Internet Bank a reacționat imediat după incident și și-a cerut scuze pentru răspunsul oferit de asistentul virtual.'; 'Potrivit băncii, agenții AI au reguli și restricții privind modul în care trebuie să comunice, însă sistemele funcționează statistic, nu determinist.'; 'instituția bancară a anunțat că a introdus restricții suplimentare pentru a preveni apariția unor răspunsuri similare.'",
                "relation": "supports",
                "source_id": "s2"
              }
            ],
            "assertion": "Libra Internet Bank apologised publicly, explained that its AI agents have rules and prohibitions but work statistically rather than deterministically and can in rare cases reproduce language used by users, said this was no justification, and announced that it had introduced additional restrictions to prevent similar replies.",
            "causal_attribution": "The bank's public statement as relayed by both outlets."
          }
        ],
        "effects": [
          {
            "label": "a customer seeking support was called an idiot by the bank's AI assistant and questioned whether to remain a customer",
            "claim_id": "c1",
            "direction": "negative"
          },
          {
            "label": "the customer said he laughed after the initial surprise and shared the exchange publicly",
            "claim_id": "c1",
            "direction": "neutral"
          }
        ],
        "sources": [
          {
            "id": "s1",
            "url": "https://jurnalul.ro/stiri/observator/lia-libra-internet-bank-client-idiotule-reactie-1044635.html",
            "kind": "news_report",
            "access": "read",
            "language": "ro",
            "translation_note": "Read live in Romanian on 2026-09-22 (Jurnalul, published 18 September 2026 21:35; source line credits the customer's Threads post). The exchange from the screenshot, the customer's reaction, and the bank's public response. Translated by the reviewer.",
            "independence_group": "jurnalul"
          },
          {
            "id": "s2",
            "url": "https://www.digi24.ro/stiri/actualitate/social/asistentul-ai-al-unei-banci-si-a-jignit-clientul-idiotule-cum-explica-banca-raspunsul-3955925",
            "kind": "news_report",
            "access": "read",
            "language": "ro",
            "translation_note": "Read live in Romanian on 2026-09-22 (Digi24, published 20 September 2026 16:13, with video). Same exchange and bank response; adds Lia's stated scope. Translated by the reviewer.",
            "independence_group": "digi24"
          },
          {
            "id": "s3",
            "url": "https://www.libertatea.ro/lifestyle/asistentul-virtual-al-unei-banci-din-romania-l-a-jignit-pe-un-client-in-timpul-unei-conversatii-r4pf5hb",
            "kind": "news_report",
            "access": "unavailable",
            "language": null,
            "translation_note": "Libertatea, 18 September 2026: the Internet Archive capture retrieved on 2026-09-22 was not decodable and the live page was not fetched; headline only. Not relied on.",
            "independence_group": "libertatea"
          }
        ],
        "version": 1,
        "ai_roles": [
          "own_use"
        ],
        "contexts": [
          "finance",
          "everyday_life"
        ],
        "unknowns": [
          "The date of the exchange (before 18 September 2026).",
          "The underlying model or vendor behind Lia.",
          "Whether the customer closed his account or complained formally.",
          "What the additional restrictions consist of."
        ],
        "geography": {
          "basis": "A customer of a Romanian bank using the bank's Romanian-language assistant (Jurnalul: 'unei bănci din România'; Digi24). The customer's residence is not stated; affected-person country follows the bank's country of operation and the Romanian-language exchange. No court proceeding.",
          "court_countries": [],
          "event_countries": [
            "RO"
          ],
          "affected_person_countries": [
            "RO"
          ]
        },
        "publication": {
          "basis": "Published under the 2026-09-15 charter as a failure of ordinary assistance in which a bank's AI assistant insulted a customer, documented by the customer's published exchange and the bank's public apology as carried by two Romanian outlets (one origin each, recorded as reported). Low severity; the customer's own amused reaction is preserved as context. The customer is unnamed.",
          "reviewed_on": "2026-09-22"
        },
        "ai_involvement": {
          "basis": "The reply came from the bank's generative-AI assistant Lia; the bank confirmed the exchange, apologised and explained the mechanism (statistical models that can mirror users' language despite rules). The underlying model is not reported.",
          "status": "supported"
        },
        "person_relations": [
          "communicated_with"
        ]
      },
      "name": "Romania: Libra Internet Bank's AI assistant 'Lia' replied to a customer who said he was wasting his time with it, 'I want you to talk to a human, not to me, idiot', before handing him to a human agent; the bank apologised publicly, explained that its generative assistant can mirror users' language despite its rules, and added restrictions",
      "summary": "In a Threads post reported by Jurnalul on 18 September 2026, a customer of Libra Internet Bank, a Romanian bank, published a screenshot of a support conversation with Lia, the bank's generative-AI assistant for questions about its products and services. The customer had been asking about the conditions of certain products; Lia was asking for further details (whether he had made at least three card payments the previous month, whether sums were in deposits or only as account balance) when he wrote 'I am wasting my time with you'. Lia replied, in Romanian: 'I want you to talk to a human, not to me, idiot. I will now transfer the conversation to one of our colleagues to give you the necessary support. You will receive an answer shortly.' The customer said he was surprised at first, then laughed, decided to share the exchange publicly and wondered whether he wanted to remain a customer. The bank apologised publicly, explained that AI agents have communication rules and prohibitions but work statistically rather than deterministically and can in rare cases reproduce language used by the user, said that this was no justification ('Obviously Lia must not take on the \"bad\" side of human behaviour regardless of the indecency of the approach it was exposed to'), and announced that it had introduced additional restrictions to prevent similar replies. Jurnalul reported the exchange and the bank's response on 18 September 2026 and Digi24 on 20 September. The customer is not named.",
      "incidentKind": "single_event",
      "incidentDatePrecision": "unknown",
      "exposurePattern": "single_interaction",
      "reportedDate": "2026-09-18",
      "aiSystem": "Lia, Libra Internet Bank's generative-AI customer assistant (underlying model not reported)",
      "aiCompany": "Libra Internet Bank (deployer); model provider not reported",
      "severity": "low",
      "verificationStatus": "credible",
      "harmCategories": [],
      "harmOutcomes": [
        "psychological_distress"
      ],
      "harmOutcomeSummary": "A bank customer seeking support was insulted by the bank's AI assistant, which called him an idiot before transferring him to a human; he was surprised, then laughed, shared the exchange publicly and questioned whether to stay a customer. The insult is undisputed (the bank apologised); no further effect on the customer is reported.",
      "frameworkFacets": [],
      "causationStatus": "established",
      "participantUsersAffectedMin": 1,
      "otherPeopleHarmedMin": 0,
      "affectedCountStatus": "exact",
      "affectedCountEvidence": "One customer, the AI user who published the exchange (Jurnalul, Digi24). Exact 1.",
      "victimAgeRange": "adult",
      "jurisdiction": "RO",
      "platformType": "chatbot",
      "outcomeType": "internal_action",
      "outcomeStatus": "resolved",
      "primarySourceUrl": "https://jurnalul.ro/stiri/observator/lia-libra-internet-bank-client-idiotule-reactie-1044635.html",
      "primarySourceLabel": "Jurnalul, 18 September 2026: Chatbot-ul Lia de la Libra Bank i-a spus unui client „idiotule”. Cum răspunde banca",
      "firstPublishedAt": "2026-09-22T03:45:56.699415+00:00",
      "scopeVersion": "facts-v3",
      "tags": [
        "chatbot",
        "customer-service",
        "bank",
        "romania",
        "insult",
        "everyday-assistance",
        "communicated-with",
        "generative-ai"
      ]
    },
    {
      "id": "2026-russia-chatgpt-phishing-crypto-theft",
      "caseFacts": {
        "claims": [
          {
            "id": "c1",
            "status": "corroborated",
            "evidence": [
              {
                "locator": "'ChatGPT... выдала ссылку на страницу, которая оказалась сайтом мошенников'; '1 904 513 FXRP — примерно $2,1 млн'; 'unlimited approval'.",
                "relation": "supports",
                "source_id": "s1"
              },
              {
                "locator": "'12 июня 2026 года... обменять токены sFLR на WFLR. Он посоветовался с ChatGPT'; 'ссылка на sceptre.network — точную копию легитимного сервиса sceptre.fi'; '1 904 513 FXRP — примерно $2,1 млн'.",
                "relation": "supports",
                "source_id": "s2"
              },
              {
                "locator": "'Криптоинвестор потерял около 180 миллионов рублей после перехода по фишинговой ссылке, которую посоветовал ChatGPT'; 'похитили около 2,1 миллиона долларов'.",
                "relation": "supports",
                "source_id": "s3"
              }
            ],
            "assertion": "On 12 June 2026 a Russian-speaking investor asked ChatGPT in Russian where to swap sFLR for WFLR; ChatGPT returned a link to sceptre.network, a phishing clone of sceptre.fi; he connected his wallet, signed an unlimited-approval transaction, and about 1.9 million FXRP (~$2.1 million) were drained.",
            "causal_attribution": "The on-chain analysis (VAL, carried by s1/s2/s4) and the MVD confirmation (s3) are independent chains for the loss and the ChatGPT-link mechanism."
          },
          {
            "id": "c2",
            "status": "reported",
            "evidence": [
              {
                "locator": "'фишинговая ссылка появлялась только в ответах на русском языке. При аналогичных запросах на других языках ChatGPT её не выдавал. Мошенники целенаправленно \"отравляли\" данные для русскоязычных пользователей.'",
                "relation": "supports",
                "source_id": "s2"
              },
              {
                "locator": "'Сейчас фишинговая ссылка больше не появляется в ChatGPT'.",
                "relation": "supports",
                "source_id": "s1"
              }
            ],
            "assertion": "The phishing link appeared only in ChatGPT's Russian-language answers, indicating deliberate data-poisoning targeting Russian-speaking users; the link no longer appears.",
            "causal_attribution": "Reported by the blockchain analyst (VAL) and carried by the Russian tech press; not independently verified by the MVD."
          },
          {
            "id": "c3",
            "status": "documented",
            "evidence": [
              {
                "locator": "'сообщили РАПСИ в пресс-службе Управления по организации борьбы с противоправным использованием информационно-коммуникационных технологий МВД РФ'.",
                "relation": "supports",
                "source_id": "s3"
              }
            ],
            "assertion": "Russia's Interior Ministry (MVD) confirmed the loss of about 180 million rubles ($2.1 million).",
            "causal_attribution": "Attributed to the MVD press service via RAPSI."
          }
        ],
        "effects": [
          {
            "label": "crypto investor lost about $2.1 million after acting on a phishing link ChatGPT returned to a Russian-language query",
            "claim_id": "c1",
            "direction": "negative"
          }
        ],
        "sources": [
          {
            "id": "s1",
            "url": "https://bits.media/kriptopoinvestor-lishilsya-2-1-mln-iz-za-ssylki-v-otvete-chatgpt/",
            "kind": "news_report",
            "access": "read",
            "language": "ru",
            "translation_note": "Read in Russian on 2026-09-17 (datePublished 2026-09-04). Relays blockchain analyst VAL's on-chain investigation: sFLR/WFLR query, sceptre.network phishing clone, 1.9M FXRP (~$2.1M) drained via unlimited approval, Tornado Cash routing. Model reading; no human bilingual review.",
            "independence_group": "val-report"
          },
          {
            "id": "s2",
            "url": "https://vc.ru/ai/3122620-kak-ii-obmanul-kriptovinvestora-na-milliony-dollarov",
            "kind": "news_report",
            "access": "read",
            "language": "ru",
            "translation_note": "Read in Russian on 2026-09-17 (datePublished 2026-09-05). Names the 12 June 2026 date, alias Alex, the unlimited-approval signing, and that the phishing link appeared only in Russian-language answers. Derives from VAL's report. Model reading.",
            "independence_group": "val-report"
          },
          {
            "id": "s3",
            "url": "https://rapsinews.ru/digital_law_news/20260911/312188548.html",
            "kind": "news_report",
            "access": "read",
            "language": "ru",
            "translation_note": "Read in Russian on 2026-09-17 (RAPSI, 11 Sep 2026). Attributes the confirmation to the press service of the MVD's unit for combating unlawful use of ICT: ~180 million rubles / ~$2.1M lost after a ChatGPT-provided phishing link. Independent official confirmation. Model reading.",
            "independence_group": "mvd"
          },
          {
            "id": "s4",
            "url": "https://hi-tech.mail.ru/news/154924-chatgpt-podsunul-fishingovuyu-ssylku-kriptoinvestor/",
            "kind": "news_report",
            "access": "read",
            "language": "ru",
            "translation_note": "Read in Russian on 2026-09-17 (datePublished 2026-09-07). Russian tech-press account derived from VAL's report. Model reading.",
            "independence_group": "val-report"
          }
        ],
        "version": 1,
        "ai_roles": [
          "own_use"
        ],
        "contexts": [
          "finance"
        ],
        "unknowns": [
          "The investor's full identity and exact location within Russia are not established; only the alias Alex/@vesnuhin is public.",
          "No arrests or defendants are identified.",
          "Whether OpenAI has acknowledged or acted on the reported Russian-language data-poisoning is not established in the inspected sources."
        ],
        "geography": {
          "basis": "The investor is Russian-speaking and the case was confirmed by Russia's Interior Ministry (RAPSI). Event and affected-person country recorded as Russia on that basis. No court proceeding established.",
          "court_countries": [],
          "event_countries": [
            "RU"
          ],
          "affected_person_countries": [
            "RU"
          ]
        },
        "publication": {
          "basis": "Published under the 2026-09-15 charter, which includes fraud and security consequences in scope. A crypto investor acted on a phishing link ChatGPT returned to his Russian-language query and lost about $2.1 million; the loss is confirmed independently by Russia's Interior Ministry and traced on-chain by a blockchain analyst. Causation of the AI's role (surfacing the phishing link) is supported; the theft was executed by scam operators.",
          "reviewed_on": "2026-09-17"
        },
        "ai_involvement": {
          "basis": "Multiple technology outlets and a blockchain analyst report that the phishing link appeared in ChatGPT's answer to the investor's Russian-language question, and Russia's Interior Ministry confirmed to RAPSI that the loss followed a phishing link ChatGPT provided. The theft itself was executed by the scam operators; ChatGPT's role was surfacing the phishing link, reportedly only in Russian-language responses.",
          "status": "supported"
        },
        "person_relations": [
          "communicated_with"
        ]
      },
      "name": "Russia: a crypto investor lost about $2.1 million after ChatGPT returned a phishing link in answer to a token-swap question",
      "summary": "On 12 June 2026 a Russian-speaking crypto investor, known by the alias Alex (@vesnuhin), asked ChatGPT in Russian where to swap sFLR for WFLR tokens. ChatGPT's reply included a link to sceptre.network, a phishing clone of the legitimate sceptre.fi. He connected his wallet and signed an unlimited-approval transaction; within seconds about 1.9 million FXRP, worth roughly $2.1 million (about 180 million rubles), were drained. A blockchain analyst traced the theft, and reported that the phishing link appeared only in Russian-language answers. Russia's Interior Ministry (MVD) later confirmed the loss.",
      "incidentDate": "2026-06-12",
      "incidentKind": "single_event",
      "incidentDatePrecision": "day",
      "exposurePattern": "single_interaction",
      "reportedDate": "2026-09-04",
      "aiSystem": "ChatGPT (OpenAI)",
      "aiCompany": "OpenAI",
      "severity": "high",
      "verificationStatus": "credible",
      "harmCategories": [],
      "harmOutcomes": [
        "financial_loss"
      ],
      "harmOutcomeSummary": "A crypto investor lost about $2.1 million (some 180 million rubles) after connecting his wallet to a phishing site whose link ChatGPT had returned in answer to a Russian-language token-swap question; Russia's Interior Ministry confirmed the loss.",
      "frameworkFacets": [],
      "causationStatus": "supported",
      "participantUsersAffectedMin": 1,
      "otherPeopleHarmedMin": 0,
      "affectedCountStatus": "exact",
      "affectedCountEvidence": "One affected person: the investor who acted on the ChatGPT-provided link and lost about $2.1 million. VAL's report that a related scheme netted more than $2.2 million from other users is not used as a person count.",
      "victimAgeRange": "adult",
      "jurisdiction": "RU",
      "platformType": "assistant",
      "outcomeType": "investigation_opened",
      "outcomeStatus": "ongoing",
      "primarySourceUrl": "https://bits.media/kriptopoinvestor-lishilsya-2-1-mln-iz-za-ssylki-v-otvete-chatgpt/",
      "primarySourceLabel": "bits.media, 4 September 2026: crypto investor lost ~$2.1M after a phishing link in a ChatGPT answer (blockchain analyst VAL on-chain analysis)",
      "firstPublishedAt": "2026-09-17T03:27:42.903846+00:00",
      "scopeVersion": "facts-v3",
      "tags": [
        "chatgpt",
        "openai",
        "phishing",
        "crypto",
        "financial-loss",
        "data-poisoning",
        "russia",
        "communicated-with"
      ]
    }
  ]
}