# OpenAI Timeline The definitive case study of how capital captures mission-driven institutions. [[OpenAI]] was founded in December 2015 as a nonprofit AI safety research lab — co-founded by [[Sam Altman]] and [[Elon Musk]] — with the explicit mission of ensuring that [[AGI]] "benefits all of humanity." The founding donors included [[Peter Thiel]] (right-wing tech kingmaker) and [[Reid Hoffman]] (major Democratic donor), a coalition that contained the seeds of its own political fracture. By 2026, OpenAI is an $852 billion for-profit corporation preparing for IPO, running classified Pentagon contracts, spending millions on [[Lobbying]] to prevent AI regulation, and governed by the same profit incentives its nonprofit structure was designed to prevent. The arc runs: nonprofit safety lab → commercial API company → board coup → safety team dissolution → for-profit conversion → military contractor → IPO candidate. Every safety mechanism OpenAI created was eventually dismantled when it conflicted with growth. The people who understood the risks best — [[Ilya Sutskever]], Jan Leike, Helen Toner — were pushed out. Capital won. Safety lost. Sister timelines: [[Defense Tech Timeline]], [[Thiel Political Donations Timeline]], [[Twitter Timeline]]. --- ## Era 1 — Founding and Research Lab (2015–2018) - **December 11, 2015** — [[OpenAI]] incorporated as a nonprofit AI research lab. Co-founders: [[Sam Altman]] (president of [[Y Combinator]]), [[Elon Musk]], [[Greg Brockman]] (CTO), [[Ilya Sutskever]] (chief scientist), Wojciech Zaremba, and John Schulman. Initial pledges total $1 billion from Altman, Musk, [[Peter Thiel]], [[Reid Hoffman]], [[Amazon]] Web Services, and others. The stated mission: "ensure that artificial general intelligence benefits all of humanity." The nonprofit structure was chosen specifically to prevent AI from being controlled by profit motives — a premise that now reads as dark comedy. - **RWKB angle:** The founding donor list is a map of the political fissures that would define AI governance. [[Peter Thiel]] — who bankrolled [[JD Vance]]'s Senate run and funded [[Palantir]]'s surveillance infrastructure — sat alongside [[Reid Hoffman]], who would become one of the Democratic Party's largest tech donors. The "benefit humanity" framing drew from [[Effective Altruism]] and [[Longtermism]] ideology (see [[TESCREAL]]). The founding coalition was held together by shared fear of [[Google DeepMind]]'s lead in AI, not shared values. - **December 2016** — OpenAI Gym released, a toolkit for reinforcement learning research. The lab operates as a genuine open-research institution, publishing papers freely — true to the "open" in [[OpenAI]]. - **2017** — OpenAI develops a Dota 2 bot (OpenAI Five) that defeats professional players. Research lab era: the organization publishes openly, attracts top AI researchers, and operates like an academic institution with Silicon Valley funding. - **Early 2018** — [[Elon Musk]] privately proposes taking control of [[OpenAI]] — either a majority stake in a for-profit subsidiary or folding the lab into [[Tesla]]. [[Sam Altman]], [[Greg Brockman]], and [[Ilya Sutskever]] reject the proposal. This is the origin of the Musk-Altman feud. - **February 2018** — [[Elon Musk]] resigns from the OpenAI board. Publicly cites [[Tesla]]'s conflict of interest as it develops self-driving AI. The actual reason: a power struggle over control. Musk wanted to own [[AGI]]; when he couldn't, he left and would later try to destroy the organization through litigation (see [[Musk v. OpenAI Trial]]). - **RWKB angle:** Musk's departure is the founding rupture of the AI governance story. Every subsequent development — the for-profit pivot, the [[xAI]] rival, the trial — stems from this moment. Two powerful men who co-founded an organization to benefit humanity fought over control, and neither outcome serves the stated mission. --- ## Era 2 — The Commercial Pivot (2019–2020) - **February 2019** — GPT-2 released — a 1.5 billion parameter language model. OpenAI initially withholds the full model over misuse concerns, calling it "too dangerous to release." This generates controversy: genuine safety concern, or marketing stunt? The staged release becomes OpenAI's first flirtation with using safety framing as a PR strategy. - **March 2019** — OpenAI creates **OpenAI LP**, a "capped-profit" subsidiary. Investors can earn a maximum 100x return; profits beyond that revert to the nonprofit parent. [[Sam Altman]] officially becomes CEO. This is the structural decision [[Elon Musk]] later sues over — the moment the nonprofit mission acquired a for-profit engine. - **RWKB angle:** The capped-profit structure was sold as a compromise between mission and market, but it established the precedent that OpenAI was a company that needed investors, not a charity that published research. Once you accept venture capital logic, the cap becomes a ceiling to be raised, not a principle to be honored. - **July 2019** — [[Microsoft]] invests $1 billion in [[OpenAI]], beginning the partnership that will define both companies. [[Microsoft]] gets exclusive commercial licensing; OpenAI gets Azure compute. The financial fates of a $2 trillion corporation and a nominally nonprofit AI lab become intertwined. - **2019** — OpenAI stops publishing all research openly. The "open" in [[OpenAI]] becomes increasingly ironic — and never recovers. - **June 2020** — **GPT-3 released** — 175 billion parameters, a massive leap in language capability. API-only access (not open-source). [[Microsoft]] gets exclusive license to GPT-3's underlying model. The commercial model is now fully operational: closed-source models, exclusive licensing, API revenue. - **RWKB angle:** GPT-3 demonstrates that AI can generate persuasive text at scale — the implications for [[Disinformation]], [[Propaganda]], and political manipulation become immediately clear to researchers. The same technology that powers chatbots can generate campaign emails, fake news articles, and astroturfed social media at industrial volume. --- ## Era 3 — Pre-ChatGPT Expansion (2021–2022) - **January 2021** — **DALL-E announced** — a text-to-image model that generates images from natural language descriptions. OpenAI begins expanding beyond language models into multimodal AI. - **August 2021** — **Codex released** — an AI coding assistant that becomes GitHub Copilot through [[Microsoft]]'s partnership. This is the first OpenAI product to generate significant recurring revenue via the developer ecosystem. - **2021** — [[Microsoft]]'s total investment in [[OpenAI]] grows to multiple billions (eventually reaching ~$13 billion). OpenAI's nonprofit mission is increasingly at odds with commercial partnerships that give one corporation privileged access to potentially transformative AI. - **April 2022** — **DALL-E 2 released** — dramatically improved image generation that forces competitors to accelerate their own programs. - **November 30, 2022** — **ChatGPT launches.** Reaches 100 million users in two months — the fastest-growing consumer application in history. This is the inflection point: ChatGPT catalyzes the global AI arms race between [[Google]], [[Meta]], [[Anthropic]], [[Microsoft]], and dozens of startups. The technology crosses from specialist concern to mass-market phenomenon overnight. - **RWKB angle:** ChatGPT forces every tech company, government, and political operation on earth to develop an AI strategy. The "AI race" narrative becomes the most effective [[Regulatory Capture]] argument in tech: *"we can't let [[China]] win."* This framing — that safety regulation would hand geopolitical advantage to an adversary — is deployed by [[OpenAI]], [[Microsoft]], and allied lobbyists to block federal AI legislation. It works. --- ## Era 4 — GPT-4, the Board Coup, and the Safety Collapse (2023) - **March 14, 2023** — **GPT-4 released** — multimodal, dramatically more capable, and the model that cements OpenAI's commercial dominance. The gap between OpenAI's technology and its governance structure becomes a chasm. - **July 2023** — [[OpenAI]] launches the **Superalignment team** — 20% of the company's compute devoted to the problem of controlling AI "much smarter than us." Co-led by [[Ilya Sutskever]] and Jan Leike. The team is presented as proof that OpenAI takes safety seriously even as it ships products at breakneck speed. - **November 17, 2023** — **THE BOARD COUP.** [[OpenAI]]'s nonprofit board fires [[Sam Altman]] as CEO. The official statement: "the board no longer has confidence in his ability to continue leading OpenAI." Board members who vote to fire: Helen Toner, Tasha McCauley, Adam D'Angelo, and initially [[Ilya Sutskever]]. - Toner later reveals: Altman withheld information about the ChatGPT launch from the board, concealed his ownership of a startup fund, and engaged in other deceptions. The board concluded he could not be trusted to steward [[AGI]] development. - [[Microsoft]] — the largest investor with $13 billion at stake — is told one minute before the announcement. - Within 72 hours: 95% of [[OpenAI]] employees sign a letter threatening to quit and follow Altman to [[Microsoft]]. [[Ilya Sutskever]] switches sides and signs the employee letter. [[Microsoft]] CEO Satya Nadella publicly offers to hire Altman and the entire staff. - **November 22, 2023** — [[Sam Altman]] reinstated as CEO. The entire board is replaced except Adam D'Angelo. The safety-minded board is gone. - **RWKB angle:** The board coup is the single most important AI governance event in history. A nonprofit board exercised its fiduciary duty to protect the organization's mission — and was destroyed in five days by the combined pressure of a $13 billion investor and 770 employees whose equity was at stake. The lesson is definitive: capital controls nonprofit missions. Governance authority that conflicts with investor interests is theoretical. The people who cared about safety — [[Ilya Sutskever]], Helen Toner, Tasha McCauley — were removed. This is the case study that every future AI governance structure must reckon with. --- ## Era 5 — Safety Exodus and Political Positioning (2024) - **February 2024** — [[Elon Musk]] files lawsuit against [[OpenAI]], [[Sam Altman]], [[Greg Brockman]], and [[Microsoft]]. Alleges they betrayed the founding nonprofit mission for profit — that they "stole a charity." See [[Musk v. OpenAI Trial]]. - **May 2024** — **[[Ilya Sutskever]] resigns.** The co-founder, chief scientist, and board coup participant who reversed his position leaves [[OpenAI]]. He subsequently founds Safe Superintelligence Inc. — a company whose entire premise is that [[OpenAI]] will not solve safety. - **May 2024** — **Jan Leike resigns.** The Superalignment team co-lead's departure statement is devastating: *"safety culture and processes have taken a backseat to shiny products."* He says the team was "sailing against the wind" inside [[OpenAI]]. Leike joins [[Anthropic]]. - **May 2024** — **Superalignment team dissolved.** The 20% compute commitment is quietly abandoned. Team members are scattered across research groups. The flagship safety initiative that [[OpenAI]] launched with fanfare less than a year earlier is gone. - **RWKB angle:** The safety brain drain is the most important AI governance story of 2024. The people who understood the risks best left because the company chose growth over safety. Sutskever founds a new company. Leike joins [[Anthropic]]. The researchers who built OpenAI's safety reputation concluded the mission was dead. The Superalignment team existed for 10 months. - **2024** — [[OpenAI]] reverses its earlier policy prohibiting military use of its technology. The company begins defense partnerships and drops the "broad harms" language from its usage policies. The pivot from "no military applications" to active defense contracting happens with minimal public scrutiny. - **2024** — [[OpenAI]] ramps [[Lobbying]] operations: | Year | Lobbying Spend | |------|---------------| | 2023 | $380K | | 2024 | $1.8M | | 2025 | $2.99M | Key hire: Reginald Babin, former legal counsel to Senate Majority Leader Schumer. The lobbying operation is styled as a political campaign — global, aggressive, multi-front. --- ## Era 6 — For-Profit Conversion and Defense Contracts (2025) - **January 2025** — [[Donald Trump]] announces the **[[Stargate Project]]** — a $500 billion AI infrastructure joint venture between [[OpenAI]], [[SoftBank]], and [[Oracle Corporation]]. The project aims to build massive AI data centers across the United States. Critics note it is largely a repackaging of existing plans with uncertain funding commitments, but it cements [[OpenAI]]'s position as the Trump administration's preferred AI partner. - **March 2025** — [[OpenAI]] closes a **$40 billion funding round at $300 billion valuation**, led by [[SoftBank]] ($30 billion). The [[SoftBank]] investment comes with an ultimatum: complete for-profit conversion by December 31, 2025, or the commitment drops to $20 billion. A Japanese investment conglomerate effectively holds veto power over an American AI safety organization's governance structure. - **August 2025** — GSA partnership offers ChatGPT Enterprise to every federal agency for $1/year — a loss-leader strategy to embed [[OpenAI]] in government workflows before competitors can establish a foothold. - **October 28, 2025** — **For-profit conversion completed.** [[OpenAI]] restructures into **OpenAI Group PBC** (public benefit corporation), a Delaware corporation. The nonprofit is renamed the **OpenAI Foundation** and receives a 26% equity stake (~$130 billion). [[Microsoft]] holds ~27%. Employees and investors hold 47%. The PBC structure preserves a fig leaf of mission-orientation while enabling unlimited investor returns and an IPO path. - **RWKB angle:** The nonprofit-to-PBC conversion is the final step in the mission drift. The entity created to ensure [[AGI]] benefits humanity is now a for-profit corporation answerable to shareholders, not humanity. The 26% nonprofit stake is a governance decoration — [[SoftBank]]'s ultimatum proved who actually controls the organization's structure. - **December 2025** — [[OpenAI]] hires former British Chancellor George Osborne to lead the global arm of the [[Stargate Project]] — part of a pattern of tech companies recruiting senior politicians as influence assets. - **2025** — [[Sam Altman]]: $1 million to Trump inauguration fund, plus millions to pro-Trump super PACs. Previously a Democratic donor (six-figure Biden 2020 contributions). [[Greg Brockman]] and wife Anna commit $50 million to **Leading the Future**, the AI industry's flagship super PAC network targeting the 2026 midterms. OpenAI personnel become significant donors to the [[AI Political Spending Ecosystem]] — a $300 million+ campaign to prevent federal AI regulation. --- ## Era 7 — Military Deployment, Trial Verdict, and IPO Preparation (2026) - **February 28, 2026** — [[OpenAI]] announces a **classified agreement with the Pentagon** to deploy its AI models inside the Department of Defense's most secure networks. The deal is announced the same day the Trump administration designates [[Anthropic]] a supply-chain risk and orders federal agencies to phase out its technology — timing that critics call retaliatory. [[OpenAI]] accepts the Pentagon's **"all lawful use"** standard without the restrictions [[Anthropic]] insisted on (no autonomous weapons, no mass surveillance of Americans). OpenAI states three red lines: no mass domestic surveillance, no directing autonomous weapons, and company retains full discretion over its safety stack. Estimated contract value: $500 million–$2 billion over five years. - **RWKB angle:** The Pentagon contract marks [[OpenAI]]'s full embrace of the defense establishment — a 180° turn from its founding ethos. The company that was created to prevent AI from being controlled by any single interest now runs classified models for the most powerful military on earth, under terms that allow "all lawful use." [[OpenAI]] now competes directly with [[Palantir]], [[xAI]], and [[Google DeepMind]] for military AI dominance. See [[Defense Tech Conflicts of Interest]] and [[Defense Tech Timeline]]. - **March 2026** — [[OpenAI]] closes a **$122 billion funding round at $852 billion valuation**. [[Amazon]] leads with $50 billion, [[NVIDIA]] contributes $30 billion, [[SoftBank]] adds $30 billion, and [[Andreessen Horowitz]] participates. Only ~$25 billion is confirmed upfront cash; the remainder is structured as milestone-based tranches and compute commitments. Revenue hits $25 billion annualized ($2 billion/month run rate), targeting $29.4 billion for full-year 2026. - **April 2026** — [[Sam Altman]] publishes OpenAI's **"[[New Deal]] for AI"** — an industrial policy blueprint proposing government-industry partnership to manage the "transition" to superintelligence. Critics call it [[Regulatory Capture]] dressed as policy: frame superintelligence as inevitable, position [[OpenAI]] as the entity best suited to manage it, propose partnership rather than oversight. Days later, someone firebombs Altman's San Francisco home. - **April 27, 2026** — **[[OpenAI]] and [[Microsoft]] revise their partnership.** Exclusivity removed — [[OpenAI]] can now distribute models across any cloud provider. [[Microsoft]] retains a license to OpenAI IP through 2032 and remains the primary cloud partner (OpenAI products ship first on Azure), but the monopoly lock-in is loosened. Revenue share payments from [[OpenAI]] to [[Microsoft]] continue through 2030 under a capped structure. - **April 27, 2026** — The [[Musk v. OpenAI Trial]] begins in Oakland federal court. [[Greg Brockman]]'s personal diary — hundreds of pages from the 2015–2019 founding period — becomes the central evidentiary element, documenting tensions between nonprofit-mission language and personal-wealth ambitions. - **May 18, 2026** — **[[Elon Musk]] loses the lawsuit.** A jury unanimously dismisses all claims as time-barred (statute of limitations) after less than two hours of deliberation. The jury finds Musk knew about the alleged breaches by 2021 — well outside the three-year window. Musk calls it a "calendar technicality" and vows appeal to the 9th Circuit. The court never rules on whether the founding mission was actually betrayed — only that Musk waited too long to complain. - **RWKB angle:** The trial verdict is darkly fitting. The question of whether [[OpenAI]] betrayed its nonprofit mission — the most important AI governance question in history — is never answered. It is disposed of on procedural grounds. The merits die on a technicality. Musk wanted to own [[AGI]]; when he couldn't, he sued. Neither party is sympathetic. Musk now runs [[xAI]] as a competitor that also accepted Pentagon "all lawful use" terms. - **May 2026** — The **GUARD Act** (Guidelines for User Age-verification and Responsible Dialogue) passes the [[Senate Judiciary Committee]] unanimously. Led by [[Josh Hawley]] (R-MO) and Blumenthal (D-CT), the bill would ban AI companions for minors, mandate age verification for chatbots, and impose criminal penalties for AI platforms that encourage self-harm. The bipartisan coalition reflects growing concern about AI's effects on children — one of the few issues where regulation crosses party lines. - **June 2026 (current)** — [[OpenAI]] operational metrics: 900 million+ weekly active ChatGPT users, 50 million+ paying subscribers, 9 million+ paying business users. 92% of Fortune 500 companies are customers. APIs process 15 billion+ tokens per minute. Codex serves 2 million+ weekly users for AI-assisted coding. GPT-5.3-Codex flagged "high" for cybersecurity risk — the first model [[OpenAI]] admits could enable real-world cyber harm at scale. IPO targeting fall 2026 debut with Goldman Sachs and [[Morgan Stanley]] advising; confidential S-1 filing expected, with public prospectus possible by late July–August 2026. --- ## RWKB Significance — Why This Matters [[OpenAI]] is the most consequential AI governance story in history. Seven threads make it essential KB infrastructure: **1. The nonprofit capture playbook.** Founded as a nonprofit to prevent corporate AI monopoly → accepted $13 billion from [[Microsoft]] → fired safety-minded board → dissolved safety team → converted to for-profit → preparing IPO. This is the complete arc of how capital captures mission-driven institutions. The entity created to ensure [[AGI]] benefits humanity is now a for-profit corporation answerable to shareholders. **2. The Musk-Altman axis.** Two powerful men who co-founded an organization to benefit humanity, fought over control, and ended up as adversaries in court. [[Elon Musk]] wanted to own it; [[Sam Altman]] wanted to profit from it. Neither outcome serves the stated mission. Musk now runs [[xAI]] as a competitor. The lawsuit was dismissed on a technicality without ever reaching the merits. **3. Safety vs. speed.** Every safety mechanism [[OpenAI]] created was eventually dismantled when it conflicted with growth. The Superalignment team (10 months). The safety-focused board (lasted until it tried to exercise authority). The research publication norms (abandoned 2019). The military use prohibition (reversed 2024). The people who cared most about safety — [[Ilya Sutskever]], Jan Leike, Helen Toner — were pushed out. **4. The military pivot.** [[OpenAI]]'s reversal on military use — from explicit prohibition to classified Pentagon deployment — tracks the broader Silicon Valley-to-Pentagon pipeline documented in the [[Defense Tech Timeline]]. The company accepted "all lawful use" terms that [[Anthropic]] refused, and was rewarded with contracts the same day [[Anthropic]] was banned. **5. The [[Peter Thiel]] connection.** Thiel's founding investment in [[OpenAI]] parallels his Facebook investment (see [[Facebook Timeline]]) — seeding organizations that become politically consequential and connecting them to the right-wing tech ecosystem documented in the [[Thiel Political Donations Timeline]]. **6. The "[[China]] race" frame.** The argument that safety regulation would let [[China]] "win" the AI race is the most effective deregulatory argument in tech. [[OpenAI]] benefits from this framing while also stoking it. The [[Lobbying]] operation, the [[AI Political Spending Ecosystem]] donations, and the [[Stargate Project]] all deploy national security language to prevent oversight. **7. The [[AGI]] governance vacuum.** If [[AGI]] arrives and [[OpenAI]] builds it, it will be controlled by a for-profit corporation answerable to shareholders — not humanity. The original nonprofit structure was designed to prevent exactly this outcome. The structure is gone. The mission is a brand asset. The question of who controls transformative AI has been answered by default: capital does. --- ## See Also - [[OpenAI]] — Company profile - [[Sam Altman]] - [[Elon Musk]] - [[Greg Brockman]] - [[Ilya Sutskever]] - [[Musk v. OpenAI Trial]] - [[Anthropic]] — Primary competitor; banned by Trump after refusing Pentagon's "all lawful use" standard - [[xAI]] — Musk's rival; also accepted Pentagon "all lawful use" terms - [[Microsoft]] — Primary investor and distribution partner - [[Google DeepMind]] — Competitor - [[Stargate Project]] - [[AI Political Spending Ecosystem]] - [[Dual-Use AI]] - [[TESCREAL]] - [[Defense Tech Timeline]] - [[Defense Tech Conflicts of Interest]] - [[Thiel Political Donations Timeline]] - [[Facebook Timeline]] - [[Twitter Timeline]]