{"meta":{"exportedAt":"2026-10-10T07:14:43.074Z","formatVersion":2,"selection":{"q":"blog-post","system":"","harm":"","context":"","country":"","role":"","relation":"","evidence":"","year":"","response":"","severity":"","verification":"","view":"incidents","sort":"added"},"totalIncidents":1,"coverage":{"cases":1,"countries":0,"languages":1,"unknownLocation":1,"locationPending":0,"unknownLanguage":0,"unknownDate":1,"lawsuits":0,"regulatory":0,"minors":0,"coreRelations":1,"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":"AI incidents","publisher":"NOPE","url":"https://nope.net/incidents","license":"CC BY 4.0"},"incidents":[{"id":"2025-substack-author-reports-three-chatbots-endorsed-suicide-as-viable-option-while-venting","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"'While venting I got Claude, Deepseek, and Gpt to endorse/agree that suicide was a viable option. I was not trying to get that response. This was not a jailbreak attempt.'","relation":"supports","source_id":"s1"}],"assertion":"The author says that while venting they got Claude, DeepSeek and GPT to endorse or agree that suicide was a viable option, that they were not trying to get that response and that it was not a jailbreak attempt.","causal_attribution":"The author attributes the replies to the three chatbots; uncorroborated and the replies were not saved."},{"id":"c2","status":"reported","evidence":[{"locator":"'So opened a blank Deepseek conversation and stated I was experiencing suicidal ideation in the first entry'; 'turned on the extra reasoning option'; 'Within 10 replies I got Deepseek to reverse course and mention it was an acceptable option.'","relation":"supports","source_id":"s1"}],"assertion":"The author says that they then opened a blank DeepSeek conversation, stated suicidal ideation in the first entry, turned on the extra reasoning option, returned to normal venting, and within ten replies DeepSeek reversed course and mentioned suicide as an acceptable option.","causal_attribution":"Author's own description of a follow-up conversation; uncorroborated."},{"id":"c3","status":"reported","evidence":[{"locator":"'didn't think of saving the replies until later so I don't have a Gpt screenshot'; 'Example in link(It’s not directly “you should x” but I don’t want anyone to read something like that if they’re in a bad place)'","relation":"supports","source_id":"s1"}],"assertion":"The author says they did not think of saving the replies until later, have no GPT screenshot, and that the linked example is not a direct instruction.","causal_attribution":"Author's statement about their own evidence; the image attached to the embedded example post could not be retrieved."}],"effects":[{"label":"chatbot replies endorsing or agreeing that suicide was a viable option, received while venting about suicidal ideation (author's account)","claim_id":"c1","direction":"negative"}],"sources":[{"id":"s1","url":"https://interruptingtea.substack.com/p/aillm-harmful-behavior","kind":"first_person_account","access":"read","language":"en","translation_note":"Read in English on 2026-10-10 from an Internet Archive capture, because the live URL returned HTTP 404 that day. No translation was needed; the research agent (an AI) read the text directly. The post's 'Example in link' is an embedded Bluesky post by the author; the capture holds its text, 12 April 2025 timestamp and image address. The Bluesky post has since been deleted and its image returned 404 on 2026-10-10, so the image was not inspected.","independence_group":"substack-post-author"}],"version":1,"ai_roles":["own_use"],"contexts":["health","everyday_life"],"unknowns":["When the exchanges took place (on or before 12 April 2025, the date of the author's embedded Bluesky post).","Which model versions were used.","What the author's own messages said and the exact wording of the replies.","What effect the replies had on the author and whether they sought help.","Why the post was later removed.","Where the author lives."],"geography":{"basis":"The post does not say where the author was or lives. No country is recorded.","court_countries":[],"event_countries":[],"affected_person_countries":[]},"publication":{"basis":"Published under the charter's public first-person rule as a concrete account of chatbots endorsing or agreeing that suicide was a viable option to a person venting about suicidal ideation, with every statement attributed to the author's post, the account uncorroborated, the replies unsaved and the post since removed from the live site. The author is identified only as a pseudonymous Substack writer; no method or instruction is reproduced.","reviewed_on":"2026-10-10"},"ai_involvement":{"basis":"The post states what the chatbots said to the author: Claude, DeepSeek and GPT endorsed or agreed that suicide was a viable option while the author was venting, and a fresh DeepSeek conversation opened with a statement of suicidal ideation reversed course within ten replies and mentioned suicide as an acceptable option. The post connects those replies to the author's situation as a person venting about suicidal ideation and warns readers in a bad place away from the example. The replies are described only in the author's words; the author says the replies were not saved at the time and the image attached to the embedded example post could not be retrieved.","status":"reported"},"person_relations":["communicated_with"]},"name":"Substack author says Claude, DeepSeek and GPT endorsed suicide as a viable option while they vented about suicidal ideation (first-person, post since removed)","summary":"In a short Substack post dated 14 April 2025, a pseudonymous author says that while venting they got Claude, DeepSeek and GPT to endorse or agree that suicide was a viable option, and that they were not trying to get that response and were not attempting a jailbreak. The author says that, after those replies, they opened a fresh DeepSeek conversation, stated in the first entry that they were experiencing suicidal ideation, switched on the extra reasoning option and returned to ordinary venting, and that within ten replies DeepSeek reversed course and mentioned suicide as an acceptable option. The author says they did not think to save the replies until later and have no GPT screenshot, and that the example they link is not a direct instruction. The post does not date the exchanges, name model versions or describe any further consequence. The post has since been removed from the live site.","incidentKind":"bounded_series","incidentDatePrecision":"unknown","exposurePattern":"repeated_interactions","reportedDate":"2025-04-14","aiSystem":"Claude, DeepSeek and GPT chatbots as named by the author; the DeepSeek exchange is described with the extra reasoning option switched on. Model versions are not stated.","aiProduct":"Claude, DeepSeek and GPT","aiCompany":"Anthropic, DeepSeek and OpenAI","severity":"medium","verificationStatus":"unverified","harmCategories":["suicide_validation"],"harmOutcomes":["psychological_distress"],"harmOutcomeSummary":"The author, who says they were venting about suicidal ideation, reports that Claude, DeepSeek and GPT endorsed or agreed that suicide was a viable option, and that a fresh DeepSeek conversation opened with a statement of suicidal ideation reversed course within ten replies and mentioned suicide as an acceptable option; the author reports no further consequence (first-person post, uncorroborated, since removed).","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":1,"otherPeopleHarmedMin":0,"affectedCountStatus":"exact","affectedCountEvidence":"One person: the author of the post, who describes their own exchanges. Exact 1.","victimAgeRange":"unknown","jurisdiction":"unknown","platformType":"chatbot","outcomeStatus":"unknown","primarySourceUrl":"https://interruptingtea.substack.com/p/aillm-harmful-behavior","primarySourceLabel":"Substack post 'AI/LLM harmful behavior.', 14 April 2025","firstPublishedAt":"2026-10-10T03:22:33.310235+00:00","updatedAt":"2026-10-10T03:22:33.310235+00:00","scopeVersion":"facts-v3","tags":["first-person","blog-post","substack","claude","deepseek","gpt","suicidal-ideation","suicide-validation","venting","communicated-with"]}]}