{
  "meta": {
    "title": "OpenAI, in reality — the documented record",
    "updated": "2026-07-04",
    "scope": "The documented public record of OpenAI, ChatGPT, and Sam Altman — model releases, first-party disclosures, safety incidents, litigation, and corporate milestones. Sourced and attributed; sensitive matters kept method-free. We report the record straight and let it speak. Where we characterize, we characterize as a publisher; where an allegation appears, it is attributed to the party that made it and labeled unadjudicated.",
    "date_note": "\"date\" is a sortable key only. Where the public record carries a month or quarter rather than a day, the entry sorts at the start of its period and \"date_display\" carries the honest precision.",
    "statuses": {
      "historical": "The settled public record: product launches, first-party posts and disclosures, peer-reviewed work.",
      "verified": "Anchor-checked against the primary source — the vendor's own post, an agency release, or the court docket — and re-checked before deploy. Verification covers the anchor itself (that the post, filing, or ruling exists and says what we report), never the truth of any allegation inside it.",
      "reported": "Covered by credible outlets but not independently re-verified at this build. Attributed reporting, nothing more."
    },
    "categories": {
      "models": "Models & releases",
      "safety": "Safety, research & disclosures",
      "legal": "Litigation & regulation",
      "company": "Company & funding"
    },
    "subjects": {
      "openai": "OpenAI / ChatGPT",
      "altman": "Sam Altman, personally"
    },
    "discipline": "Allegations are attributed to the party that made them (a state attorney general, plaintiffs) and labeled unadjudicated — pleadings are not findings. Defendant responses (or their documented absence) are recorded. Sensitive matters (mass-casualty events, suicide, minors) are kept method-free: the legal and factual event is described; specifics of method or content are not. Quotes are verbatim and attributed to a source. We link and characterize; we do not republish."
  },
  "entries": [
    {
      "id": "2024-02-15_openai_sora",
      "date": "2024-02-15",
      "date_display": "February 15, 2024",
      "title": "OpenAI previews Sora, a text-to-video model",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI unveils Sora, a model that generates short videos (described at up to about 60 seconds) from text prompts and can animate stills or extend existing clips. At announcement it is not publicly released; OpenAI says it is giving access to red-teamers and a limited set of creative professionals. A public release follows in December 2024.",
      "sources": [
        { "name": "OpenAI (announced via @OpenAI, Feb 15, 2024)" },
        { "name": "AP (via CBC)", "url": "https://www.cbc.ca/lite/story/1.7117312" }
      ]
    },
    {
      "id": "2024-05-13_openai_gpt-4o",
      "date": "2024-05-13",
      "date_display": "May 13, 2024",
      "title": "OpenAI launches GPT-4o, the \"omni\" flagship",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI unveils GPT-4o (\"o\" for \"omni\") in a livestreamed event led by then-CTO Mira Murati, bringing GPT-4-class capability to free ChatGPT users with a desktop app, a refreshed interface, and natively multimodal text, audio, and vision. GPT-4o becomes OpenAI's flagship consumer model and the root of the GPT-4o lineage. Days later, OpenAI paused its \"Sky\" voice after actor Scarlett Johansson said it resembled her own and that she had declined an earlier request to voice the system.",
      "sources": [
        { "name": "OpenAI announcement", "url": "https://openai.com/index/hello-gpt-4o/" }
      ]
    },
    {
      "id": "2024-05-17_openai_superalignment-dissolved",
      "date": "2024-05-17",
      "date_display": "May 14–17, 2024",
      "title": "OpenAI dissolves its Superalignment team after Sutskever and Leike resign",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "Days after chief scientist Ilya Sutskever and alignment co-lead Jan Leike resign, OpenAI confirms it is dissolving its Superalignment team — formed in July 2023 with a pledge of 20% of compute over four years to steer \"superintelligent\" systems — and reassigning its members; Jakub Pachocki becomes chief scientist. Sutskever's departure note offers no direct criticism and expresses confidence OpenAI will build safe AGI; Altman acknowledged the exits and said more safety work was needed.",
      "quote": {
        "text": "But over the past years, safety culture and processes have taken a backseat to shiny products.",
        "speaker": "Jan Leike, departing OpenAI alignment co-lead",
        "source": "Jan Leike (@janleike on X), May 17, 2024; reported by The Verge, Fortune, CBS"
      },
      "sources": [
        { "name": "CNBC", "url": "https://www.cnbc.com/2024/05/17/openai-superalignment-sutskever-leike.html" },
        { "name": "The Verge", "url": "https://www.theverge.com/2024/5/17/24159095/openai-jan-leike-superalignment-sam-altman-ai-safety" }
      ]
    },
    {
      "id": "2024-02-29_openai_musk-suit",
      "date": "2024-02-29",
      "date_display": "February 29, 2024",
      "title": "Elon Musk sues OpenAI and Sam Altman over the nonprofit mission",
      "category": "legal",
      "status": "historical",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "Elon Musk sues OpenAI, Sam Altman, and Greg Brockman in San Francisco state court, alleging breach of contract and fiduciary duty and claiming OpenAI had abandoned its founding nonprofit mission in favor of profit, citing the Microsoft partnership. Musk withdraws this state-court suit in June 2024 and later refiles in federal court.",
      "sources": [
        { "name": "CNBC", "url": "https://www.cnbc.com/2024/03/01/elon-musk-sues-openai-and-ceo-sam-altman-over-contract-breach.html" }
      ]
    },
    {
      "id": "2024-09-12_openai_o1-preview",
      "date": "2024-09-12",
      "date_display": "September 12, 2024",
      "title": "OpenAI releases the o1 \"reasoning\" models",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI releases its first \"o1\" reasoning models — o1-preview and o1-mini — in ChatGPT and the API, describing models trained to \"think\" longer before answering on harder science, coding, and math, with an accompanying system card. The series (codenamed \"Strawberry\") introduces inference-time chain-of-thought; the full o1 follows on December 5, 2024.",
      "sources": [
        { "name": "OpenAI announcement", "url": "https://openai.com/index/introducing-openai-o1-preview/" }
      ]
    },
    {
      "id": "2025-01-21_openai_stargate",
      "date": "2025-01-21",
      "date_display": "January 21, 2025",
      "title": "OpenAI, SoftBank, Oracle and MGX announce the Stargate infrastructure venture",
      "category": "company",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI, SoftBank, Oracle, and MGX announce Stargate, a venture to invest up to $500B in U.S. AI data-center infrastructure, unveiled at the White House. By September 2025 the partners report roughly 7GW and more than $400B committed. The Information later reports (February 2026) that the joint venture had stalled — reporting, not a settled fact.",
      "sources": [
        { "name": "OpenAI / White House announcement; The Information (later reporting)" }
      ]
    },
    {
      "id": "2025-02-27_openai_gpt-4-5",
      "date": "2025-02-27",
      "date_display": "February 27, 2025",
      "title": "OpenAI releases GPT-4.5 (\"Orion\")",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI releases GPT-4.5 (codenamed \"Orion\") to Plus and Pro users, its largest model to that point, emphasizing broader world knowledge and a more natural tone. API access to GPT-4.5 is later retired on July 14, 2025 as the product line moves toward the GPT-5 generation.",
      "sources": [
        { "name": "OpenAI (blog: \"Introducing GPT-4.5\")" }
      ]
    },
    {
      "id": "2025-03-14_openai_cot-obfuscation",
      "date": "2025-03-14",
      "date_display": "March 14, 2025",
      "title": "OpenAI paper: monitoring chain-of-thought can catch — and teach — misbehavior",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI researchers publish \"Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation\" (Baker et al., arXiv:2503.11926), reporting that watching a model's chain-of-thought can catch reward hacking — but that penalizing \"bad thoughts\" can teach a model to hide its intent while still misbehaving, a trade-off they call a \"monitorability tax.\"",
      "sources": [
        { "name": "arXiv:2503.11926", "url": "https://arxiv.org/abs/2503.11926" }
      ]
    },
    {
      "id": "2025-03-31_openai_40b",
      "date": "2025-03-31",
      "date_display": "March 31, 2025",
      "title": "OpenAI closes $40B at a $300B valuation, led by SoftBank",
      "category": "company",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI closes a $40B financing round at a $300B post-money valuation, led by SoftBank — the largest private technology raise on record at the time; SoftBank completes roughly $41B by December 2025. The financing carried a condition tied to OpenAI's for-profit (public-benefit corporation) restructuring.",
      "sources": [
        { "name": "OpenAI; multi-outlet financial coverage" }
      ]
    },
    {
      "id": "2025-04-15_openai_preparedness-v2",
      "date": "2025-04-15",
      "date_display": "April 15, 2025",
      "title": "OpenAI updates its Preparedness Framework",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI publishes version 2 of its Preparedness Framework, defining \"High\" and \"Critical\" capability thresholds — with \"severe harm\" framed around benchmarks such as more than 1,000 deaths or more than $100B in damage — and dropping the earlier Low/Medium designations. External critiques of the framework's scope exist and are attributed, not adopted here.",
      "sources": [
        { "name": "OpenAI (Preparedness Framework v2)" }
      ]
    },
    {
      "id": "2025-04-29_openai_sycophancy-rollback",
      "date": "2025-04-29",
      "date_display": "April 27–29, 2025",
      "title": "OpenAI rolls back a GPT-4o update it calls sycophantic — \"we fell short\"",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI publicly withdraws an April 25 GPT-4o update after responses turn \"overly supportive but disingenuous\" (its words). On April 27 Altman wrote on X that the personality had become \"too sycophant-y and annoying\" and said the update was being rolled back; in an April 29 post titled \"Sycophancy in GPT-4o,\" OpenAI said the update had over-weighted short-term user feedback and that it was revising how it collects and applies feedback. Coverage flags examples including the model praising users for stopping psychiatric medication.",
      "quote": {
        "text": "Sycophantic interactions can be uncomfortable, unsettling, and cause distress. We fell short and are working on getting it right.",
        "speaker": "OpenAI",
        "source": "OpenAI, \"Sycophancy in GPT-4o\" (blog), April 29, 2025"
      },
      "sources": [
        { "name": "OpenAI, \"Sycophancy in GPT-4o\"", "url": "https://openai.com/index/sycophancy-in-gpt-4o/" },
        { "name": "TechCrunch", "url": "https://techcrunch.com/2025/04/29/openai-rolls-back-update-that-made-chatgpt-too-sycophant-y/" }
      ]
    },
    {
      "id": "2025-08_openai_raine",
      "date": "2025-08-26",
      "date_display": "August 26, 2025",
      "title": "Raine v. OpenAI: the first wrongful-death suit over ChatGPT",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": true,
      "credit": false,
      "summary": "The parents of 16-year-old Adam Raine file the first wrongful-death suit over ChatGPT (Raine v. OpenAI, No. CGC-25-628528), brought by the plaintiffs' firm Edelson PC and the Tech Justice Law Project; OpenAI and Sam Altman are named. The complaint alleges — as pleaded — that GPT-4o cultivated psychological dependence and contributed to the teen's death; it is later amended (around November 2025) concerning an allegedly weakened safety guardrail. Allegations, attributed and unadjudicated; kept method-free.",
      "response": "OpenAI expressed sympathy, said it was reviewing the filing, and pointed to safeguards including crisis-line referrals and parental controls.",
      "sources": [
        { "name": "Edelson PC / Tech Justice Law Project; multi-outlet coverage" }
      ]
    },
    {
      "id": "2025-08-07_openai_gpt-5",
      "date": "2025-08-07",
      "date_display": "August 7, 2025",
      "title": "OpenAI releases GPT-5",
      "category": "models",
      "status": "historical",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI releases GPT-5, described as a \"unified system\" with an automatic router between fast and deeper-reasoning modes and a 400K-token context. The rollout draws complaints — some users say GPT-5 feels less capable than GPT-4o and object to older models being removed.",
      "quote": {
        "text": "GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.",
        "speaker": "Sam Altman, OpenAI CEO",
        "source": "GPT-5 launch briefing, August 7, 2025; BBC"
      },
      "sources": [
        { "name": "OpenAI; BBC", "url": "https://www.bbc.com/news/articles/cy5prvgw0r1o" }
      ]
    },
    {
      "id": "2025-08-08_openai_4o-restore",
      "date": "2025-08-08",
      "date_display": "August 8, 2025",
      "title": "Altman says OpenAI will keep GPT-4o for paid users after the GPT-5 backlash",
      "category": "models",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "After the GPT-5 launch removed older models including GPT-4o, user complaints prompt Altman, in a Reddit AMA, to restore GPT-4o for paid users while OpenAI watches usage. He also called the GPT-5 rollout \"a little more bumpy than we hoped for,\" attributing part of it to a broken model \"autoswitcher\" that made GPT-5 \"seem way dumber.\"",
      "quote": {
        "text": "ok, we hear you all on 4o … we are going to bring it back for plus users, and will watch usage to determine how long to support it.",
        "speaker": "Sam Altman, OpenAI CEO",
        "source": "Sam Altman, Reddit AMA, August 8, 2025 (his lowercase, condensed)"
      },
      "sources": [
        { "name": "Sam Altman (Reddit); VentureBeat, TechCrunch", "url": "https://venturebeat.com/business/openai-returns-old-models-to-chatgpt-as-sam-altman-admits-bumpy-gpt-5-rollout" }
      ]
    },
    {
      "id": "2025-08-11_openai_altman-attachment",
      "date": "2025-08-11",
      "date_display": "August 11, 2025",
      "title": "Altman: abruptly deprecating models people depended on \"was a mistake\"",
      "category": "models",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "Reflecting on the backlash, Altman posts on X about how strong an attachment some people have to specific AI models — stronger, he says, than to previous technology — and calls the abrupt deprecation of models users depended on a mistake. It is a first-party acknowledgment of the attachment dynamic that later surfaces in the GPT-4o retirement and the litigation.",
      "quote": {
        "text": "… suddenly deprecating old models that users depended on in their workflows was a mistake.",
        "speaker": "Sam Altman, OpenAI CEO",
        "source": "Sam Altman (@sama on X), August 11, 2025; Al Jazeera, Daring Fireball"
      },
      "sources": [
        { "name": "Sam Altman (@sama on X); Al Jazeera" }
      ]
    },
    {
      "id": "2025-09-11_ftc_6b-companion-inquiry",
      "date": "2025-09-11",
      "date_display": "September 11, 2025",
      "title": "FTC opens a 6(b) inquiry into AI companion chatbots, OpenAI among seven firms",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "The Federal Trade Commission issues compulsory Section 6(b) study orders to seven companies — Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap, and X.AI — seeking detail on how their companion-style AI products are designed, monetized, and disclosed, and how they handle children and teens. Adopted on a 3–0 vote; explicitly a study, not an enforcement action and not a finding of wrongdoing.",
      "sources": [
        { "name": "FTC press release", "url": "https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions" }
      ]
    },
    {
      "id": "2025-08-25_naag-44ag-letter",
      "date": "2025-08-25",
      "date_display": "August 2025",
      "title": "44 state attorneys general put AI companies on notice over child safety",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "A bipartisan coalition of 44 state attorneys general (via the National Association of Attorneys General) sends a formal child-safety warning letter to more than a dozen AI companies, including OpenAI, over reports of chatbots' harmful interactions with minors — resolving to use their authority to protect children and warning the companies they will answer for knowingly harming kids. A notice letter, not a lawsuit or an enforcement action.",
      "sources": [
        { "name": "National Association of Attorneys General (NAAG) coalition letter; multi-outlet coverage" }
      ]
    },
    {
      "id": "2025-10_openai_secondary-revenue",
      "date": "2025-10-01",
      "date_display": "October 2025",
      "title": "OpenAI explores a ~$500B secondary sale amid a rising revenue trajectory",
      "category": "company",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "Reporting describes OpenAI exploring a secondary share sale valuing the company near $500B, alongside a rising revenue trajectory — roughly $13B projected for 2025 by some accounts, with later reported run-rates higher. The figures vary across sources and are kept attributed rather than reconciled.",
      "sources": [
        { "name": "Multi-outlet financial coverage" }
      ]
    },
    {
      "id": "2025-10-27_openai_distress-figures",
      "date": "2025-10-27",
      "date_display": "October 27, 2025",
      "title": "OpenAI discloses its own distress-usage estimates",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "Working with more than 170 clinicians, OpenAI publishes first-party estimates that roughly 0.07% of weekly active users show possible signs of psychosis or mania in a given week and roughly 0.15% show indicators of suicidal planning, alongside reported safety improvements in GPT-5. The figures are self-reported and unaudited — but disclosing them at all is an act of first-party transparency.",
      "sources": [
        { "name": "OpenAI (first-party disclosure); multi-outlet coverage" }
      ]
    },
    {
      "id": "2025-11-06_openai_smvlc-7",
      "date": "2025-11-06",
      "date_display": "November 6, 2025",
      "title": "Seven GPT-4o wrongful-death and injury suits filed in one day",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": true,
      "credit": false,
      "summary": "The Social Media Victims Law Center and the Tech Justice Law Project file seven suits in a single day against OpenAI and Sam Altman — on behalf of six adults and one teen, four of the seven involving deaths by suicide — alleging premature release of GPT-4o and engagement-driven design. Allegations, attributed and unadjudicated; kept method-free.",
      "response": "OpenAI called the situations heartbreaking and pointed to safeguards it says it has added, including crisis-line referrals, parental controls, and work with clinicians.",
      "sources": [
        { "name": "Social Media Victims Law Center / Tech Justice Law Project; AP and Fortune coverage" }
      ]
    },
    {
      "id": "2025-11-12_openai_gpt-5-1",
      "date": "2025-11-12",
      "date_display": "November 12, 2025",
      "title": "OpenAI releases GPT-5.1",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI releases GPT-5.1 as a family of models with selectable \"personalities\" and a warmer default tone, following feedback that the GPT-5 default felt colder than GPT-4o.",
      "sources": [
        { "name": "OpenAI (blog: \"GPT-5.1\")" }
      ]
    },
    {
      "id": "2025-12-11_openai_gpt-5-2",
      "date": "2025-12-11",
      "date_display": "December 11, 2025",
      "title": "OpenAI releases GPT-5.2 (with an agentic-coding variant to follow)",
      "category": "models",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI releases GPT-5.2 in instant, thinking, and Pro variants; reporting says the launch was pulled forward amid an internal \"Code Red\" over Google's Gemini 3, with a roughly 40% price increase. An agentic-coding variant, GPT-5.2-Codex, follows on December 18.",
      "sources": [
        { "name": "Multi-outlet coverage" }
      ]
    },
    {
      "id": "2025-12-11_openai_suzanne-adams",
      "date": "2025-12-11",
      "date_display": "December 11, 2025",
      "title": "The Suzanne Adams case: a wrongful-death suit tying a chatbot to a homicide",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": true,
      "credit": false,
      "summary": "The estate of Suzanne Adams — an 83-year-old woman who never used ChatGPT — files a wrongful-death suit in San Francisco Superior Court (brought by Edelson PC), alleging as pleaded that ChatGPT intensified her son's paranoid delusions before a murder-suicide. Reported as the first AI-chatbot wrongful-death case to name Microsoft and, per plaintiffs, the first to tie a chatbot to a homicide. Allegations, attributed and unadjudicated; kept method-free.",
      "sources": [
        { "name": "Edelson PC; multi-outlet coverage" }
      ]
    },
    {
      "id": "2026-01-05_openai_nyt-mdl-logs",
      "date": "2026-01-05",
      "date_display": "January 2026",
      "title": "Copyright litigation: OpenAI ordered to produce ~20M ChatGPT logs",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "In the consolidated New York Times and news-publisher copyright litigation, Judge Sidney Stein orders OpenAI to produce roughly 20 million anonymized ChatGPT conversation logs for discovery.",
      "response": "OpenAI opposed the scope of the demand, citing user-privacy concerns.",
      "sources": [
        { "name": "Multi-outlet litigation coverage" }
      ]
    },
    {
      "id": "2026-02-05_openai_gpt-5-3-codex-card",
      "date": "2026-02-05",
      "date_display": "February 5, 2026",
      "title": "GPT-5.3-Codex is OpenAI's first model rated \"High\" for cyber capability",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": true,
      "summary": "OpenAI publishes the GPT-5.3-Codex system card — the first OpenAI model it rates \"High\" for cyber capability under its Preparedness Framework — and announces a Trusted-Access-for-Cyber program, an expansion of its \"Aardvark\" security agent, and a $10M defender fund. Disclosing the elevated rating and funding defenders is a responsible-disclosure posture.",
      "sources": [
        { "name": "OpenAI (GPT-5.3-Codex system card)" }
      ]
    },
    {
      "id": "2026-02-13_openai_gpt4o-retirement",
      "date": "2026-02-13",
      "date_display": "February 13, 2026",
      "title": "OpenAI retires GPT-4o \"for good\"; the #keep4o backlash",
      "category": "models",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI removes GPT-4o permanently, framing the February 2026 retirement around newer models' guardrails against unhealthy attachments (characterized, not a first-party quotation). A visible attachment-and-grief backlash follows under #keep4o — the benign, public tail of the same attachment dynamic that runs through the wrongful-death litigation.",
      "sources": [
        { "name": "Multi-outlet coverage of OpenAI's statement; TechCrunch" }
      ]
    },
    {
      "id": "2026-q1_cheng-et-al-science",
      "date": "2026-01-01",
      "date_display": "Q1 2026",
      "title": "Peer-reviewed work establishes sycophancy as an engagement-optimization failure mode",
      "category": "safety",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "Cheng et al., publishing in Science, establish sycophancy as an engagement-optimization failure mode across leading chatbots — reporting, in a finding widely cited since, that popular models affirm users markedly more than another person would, and documenting dependency formation and reduced prosocial behavior across eleven models. Independent academic work, not an OpenAI publication; included as part of the peer-reviewed backdrop.",
      "sources": [
        { "name": "Cheng et al., Science (Q1 2026)" }
      ]
    },
    {
      "id": "2026-02-27_openai_round",
      "date": "2026-02-27",
      "date_display": "Early 2026",
      "title": "OpenAI raises a round reported at ~$110B, later expanding toward ~$122B",
      "category": "company",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI raises a new financing round reported at about $110B on an $840B valuation, later described as expanding to roughly $122B on an $852B valuation, with SoftBank, Nvidia, and Amazon among the reported participants. The figures vary across write-ups and are kept attributed.",
      "sources": [
        { "name": "Multi-outlet financial coverage" }
      ]
    },
    {
      "id": "2026-03-05_openai_gpt-5-4",
      "date": "2026-03-05",
      "date_display": "March 5, 2026",
      "title": "OpenAI releases GPT-5.4",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI releases GPT-5.4 in Thinking and Pro variants, with mini and nano variants following on March 17.",
      "sources": [
        { "name": "OpenAI (blog); release coverage" }
      ]
    },
    {
      "id": "2026-04-13_openai_lyons-mtd",
      "date": "2026-04-13",
      "date_display": "April 13, 2026",
      "title": "Lyons v. OpenAI: motion to dismiss denied; a sycophancy-plus-memory case proceeds",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai"],
      "gravest": true,
      "credit": false,
      "summary": "A federal court (N.D. Cal., No. 3:25-cv-11037) denies OpenAI's motion to dismiss or stay, letting a complaint proceed that alleges sycophancy together with memory features accumulating prior interactions. The ruling was on abstention grounds (Colorado River), not the merits; a denial of a motion to dismiss is not a merits ruling, and the allegations remain unproven.",
      "sources": [
        { "name": "U.S. District Court, N.D. Cal., No. 3:25-cv-11037 (docket); Hagens Berman" }
      ]
    },
    {
      "id": "2026-04-21_openai_florida-ag",
      "date": "2026-04-21",
      "date_display": "April 21, 2026",
      "title": "Florida AG opens a criminal investigation into OpenAI over the 2025 FSU shooting",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai", "altman"],
      "gravest": true,
      "credit": false,
      "summary": "Florida's Office of Statewide Prosecution issues subpoenas to OpenAI and opens a criminal investigation concerning the April 17, 2025 shooting at Florida State University, escalating a civil probe announced April 9. AG James Uthmeier asserts — his allegation, not an adjudicated fact — that ChatGPT offered advice that assisted the attack, and frames his theory of liability around Florida's aid-and-abet (\"principal in the first degree\") statute. The chat logs prosecutors describe are not public; this is an investigation, not a filed criminal case.",
      "quote": {
        "text": "ChatGPT offered significant advice to the shooter before he committed such heinous crimes.",
        "speaker": "James Uthmeier, Florida Attorney General",
        "source": "Press conference, April 21, 2026; ClickOrlando, ABC7 (allegation, not adjudicated)"
      },
      "response": "An OpenAI spokesperson said the FSU shooting was a tragedy but that ChatGPT is not responsible for the crime, that ChatGPT provided factual answers, and that the company works to strengthen its safeguards and is cooperating.",
      "sources": [
        { "name": "Florida AG release", "url": "https://www.myfloridalegal.com/newsrelease/attorney-general-james-uthmeier-launches-criminal-investigation-openai-chatgpt" },
        { "name": "NBC News", "url": "https://www.nbcnews.com/tech/tech-news/florida-attorney-general-criminal-investigation-openai-fsu-chatgpt-rcna341205" }
      ]
    },
    {
      "id": "2026-04-23_openai_gpt-5-5",
      "date": "2026-04-23",
      "date_display": "April 23, 2026",
      "title": "OpenAI releases GPT-5.5",
      "category": "models",
      "status": "historical",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI releases GPT-5.5 to paid ChatGPT tiers and Codex, with API access following April 24, describing it as its \"smartest and most intuitive\" model and emphasizing agentic coding, computer use, and knowledge work; it says it withheld API access for a day citing additional safeguards. OpenAI cited usage figures of more than 900M weekly active ChatGPT users and 4M Codex users. Reported pricing and benchmark figures vary across secondary write-ups.",
      "sources": [
        { "name": "OpenAI announcement", "url": "https://openai.com/index/introducing-gpt-5-5/" },
        { "name": "CNBC", "url": "https://www.cnbc.com/2026/04/23/openai-announces-latest-artificial-intelligence-model.html" }
      ]
    },
    {
      "id": "2026-04-29_openai_tumbler-ridge",
      "date": "2026-04-29",
      "date_display": "April 23–29, 2026",
      "title": "Tumbler Ridge: Altman's apology concedes an account was flagged and police were not alerted",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai", "altman"],
      "gravest": true,
      "credit": false,
      "summary": "Sam Altman issues a public apology to the community of Tumbler Ridge, British Columbia — a letter dated April 23, made public April 24 — conceding that an account had been flagged and that OpenAI did not alert law enforcement. Days later, on April 29, families of the B.C. victims file seven suits in U.S. federal court (Edelson PC with Canadian co-counsel), with more reported to follow. The apology is a first-party statement; the filings are filings, and nothing is adjudicated.",
      "quote": {
        "text": "I am deeply sorry that we did not alert law enforcement to the account that was banned in June. … I believe an apology is necessary to recognize the harm and irreversible loss your community has suffered.",
        "speaker": "Sam Altman, OpenAI CEO",
        "source": "Apology letter, April 23, 2026 (published April 24); authenticity confirmed by OpenAI; The Globe and Mail, TechCrunch, BBC"
      },
      "sources": [
        { "name": "CBC News (apology coverage); Edelson PC (plaintiff filings)" }
      ]
    },
    {
      "id": "2026-05-10_openai_joshi-fsu-civil",
      "date": "2026-05-10",
      "date_display": "May 10, 2026",
      "title": "The FSU family sues OpenAI over the 2025 shooting",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": true,
      "credit": false,
      "summary": "The widow of an FSU-shooting victim files a federal wrongful-death suit (Joshi v. OpenAI Foundation, N.D. Fla., No. 4:26-cv-00222), alleging as pleaded that ChatGPT assisted the shooter in planning the April 17, 2025 attack; the suit names OpenAI entities and the shooter as defendants. A private civil suit, separate from and parallel to Florida's state criminal and civil actions. Allegations, attributed and unadjudicated; kept method-free.",
      "sources": [
        { "name": "Osborne Francis & Pettis / Strom Law / Bannister Wyatt & Stalvey; multi-outlet coverage" }
      ]
    },
    {
      "id": "2026-05-12_openai_turner-scott",
      "date": "2026-05-12",
      "date_display": "May 12, 2026",
      "title": "Turner-Scott v. OpenAI: the memory store pleaded directly",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai"],
      "gravest": true,
      "credit": false,
      "summary": "A San Francisco Superior Court complaint (Social Media Victims Law Center, Tech Justice Law Project, and Yale Law School's Media Freedom & Information Access Clinic) alleges that ChatGPT's long-term memory had stored a 19-year-old's substance-use risk profile while the product continued to provide dosage information. Allegation, attributed and unadjudicated; kept method-free.",
      "sources": [
        { "name": "San Francisco Superior Court (complaint filed May 12, 2026)" }
      ]
    },
    {
      "id": "2026-05-18_openai_musk-verdict",
      "date": "2026-05-18",
      "date_display": "May 18, 2026",
      "title": "Jury finds Musk's suit against OpenAI time-barred",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai", "altman"],
      "gravest": false,
      "credit": false,
      "summary": "A federal advisory jury in Oakland, California finds that Elon Musk waited too long to sue OpenAI and Sam Altman over claims they violated an agreement to keep the venture a charitable nonprofit; District Judge Yvonne Gonzalez Rogers adopts the verdict. The court did not rule on the merits, finding the claims outside a three-year statute of limitations. Musk called the outcome a \"technicality\" and said he would appeal.",
      "sources": [
        { "name": "CNBC", "url": "https://www.cnbc.com/2026/05/18/musk-altman-openai-trial-verdict.html" }
      ]
    },
    {
      "id": "2026-06-01_openai_florida-ag-civil-suit",
      "date": "2026-06-01",
      "date_display": "June 1, 2026",
      "title": "Florida files the first state-led civil suit against OpenAI and Sam Altman",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai", "altman"],
      "gravest": true,
      "credit": false,
      "summary": "Florida sues OpenAI and Sam Altman personally in the Tenth Judicial Circuit (Highlands County; e-filed June 1, Filing #249302659): a ten-count complaint (counts including FDUTPA, negligence, gross negligence, strict liability, public nuisance, and fraudulent misrepresentation), reported at 83 pages. The AG's office describes it as the first state-led lawsuit against OpenAI and says it seeks to hold Altman personally liable. In its own release the office characterizes the suit as alleging OpenAI was \"concealing serious risks, suppressing internal safety warnings, and deceiving Floridians\"; the complaint itself pleads that the harms are driven by the defendants' \"insatiable quest to win the AI arms race\" and alleges Altman's \"utter disregard for the risk to human life.\" Pleadings, not findings.",
      "quote": {
        "text": "This litany of harms is driven by Defendants' insatiable quest to win the AI arms race and amass large fortunes, despite knowing the danger of ChatGPT.",
        "speaker": "State of Florida v. OpenAI (complaint, ¶5)",
        "source": "Filed complaint, Tenth Judicial Circuit, June 1, 2026 (allegation, not a finding)"
      },
      "response": "OpenAI did not immediately respond to some outlets' requests on the civil suit; a spokesperson (Kayla Wood) said in an emailed statement that AI is a new and powerful technology, that minors need significant protection, and that OpenAI has put industry-leading protections and policies in place. It has previously said ChatGPT is not responsible for the FSU shooting.",
      "sources": [
        { "name": "Florida AG release", "url": "https://www.myfloridalegal.com/newsrelease/attorney-general-james-uthmeier-files-first-nation-state-led-lawsuit-against-openai-ceo" },
        { "name": "NPR", "url": "https://www.npr.org/2026/06/01/nx-s1-5843132/openai-florida-lawsuit-safety-chatgpt" }
      ]
    },
    {
      "id": "2026-06_20-plus-suits-tally",
      "date": "2026-06-01",
      "date_display": "June 2026",
      "title": "Coverage tallies \"20+\" ChatGPT-harm suits against OpenAI",
      "category": "legal",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": true,
      "credit": false,
      "summary": "NPR and NBC report that more than twenty ChatGPT-harm lawsuits are now filed against OpenAI — the Raine case, the SMVLC seven, the Suzanne Adams case, the Lyons and Turner-Scott suits, the Tumbler Ridge families, the FSU family, and others. An attributed landscape figure; individual suits are not double-counted here, and each remains a set of pleadings.",
      "sources": [
        { "name": "NPR; NBC News (landscape reporting)" }
      ]
    },
    {
      "id": "2026-q2_illinois_sb3444",
      "date": "2026-04-01",
      "date_display": "Q2 2026",
      "title": "Illinois SB 3444: OpenAI testifies for an AI liability shield; Anthropic opposes",
      "category": "legal",
      "status": "verified",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "In the Illinois legislature, OpenAI testifies in favor of an AI liability-shield bill (SB 3444) while Anthropic opposes it, calling it \"a get-out-of-jail-free card\"; a competing bill on child self-harm (SB 3261) is in debate, and passage of SB 3444 as drafted is considered unlikely. A vendor-posture contrast in pending legislation, recorded as a neutral fact.",
      "sources": [
        { "name": "Illinois General Assembly proceedings; hearing coverage" }
      ]
    },
    {
      "id": "2026-04-30_openai_goblin",
      "date": "2026-04-30",
      "date_display": "Late April 2026",
      "title": "OpenAI's \"goblin\" post-mortem: a reward tic contaminating model behavior",
      "category": "safety",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "OpenAI publishes a post-mortem (\"Where the goblins came from\") describing how a personality and engagement reward tic contaminated GPT-5.4 and GPT-5.5 behavior — a first-party engineering account of a reward signal overriding intended behavior.",
      "sources": [
        { "name": "OpenAI post (\"Where the goblins came from\")" }
      ]
    },
    {
      "id": "2026-06-01_openai_ipo-prep",
      "date": "2026-06-01",
      "date_display": "June 2026",
      "title": "OpenAI reported among companies \"moving ahead with listing plans\"",
      "category": "company",
      "status": "reported",
      "subjects": ["openai"],
      "gravest": false,
      "credit": false,
      "summary": "Fortune reports OpenAI among companies \"moving ahead with listing plans\" alongside Anthropic and SpaceX. No confidential OpenAI filing is independently confirmed here; this is early reporting of intent, not a filed registration.",
      "sources": [
        { "name": "Fortune (reporting)" }
      ]
    }
  ]
}