# Anthropic Timeline The counter-narrative to every other tech company timeline in this vault. [[Anthropic]] was founded in January 2021 by [[Dario Amodei]] and [[Daniela Amodei]] — along with five other ex-[[OpenAI]] researchers — as a direct response to [[OpenAI]]'s commercialization trajectory. The founding premise: the safety researchers who couldn't change [[OpenAI]] from inside left to build an alternative. Incorporated as a Public Benefit Corporation from day one, [[Anthropic]] explicitly chose a structure that balances profit with public benefit — a structural commitment that [[OpenAI]]'s nonprofit governance was supposed to provide but ultimately abandoned (see [[OpenAI Timeline]]). The arc runs: safety schism at [[OpenAI]] → independent lab → [[Constitutional AI]] methodology → Claude model family → [[Amazon]] and [[Google]] megadeals → Pentagon standoff with [[Donald Trump]] → federal lawsuits → IPO filing. Where [[OpenAI]]'s timeline is a story of how capital captures mission, [[Anthropic]]'s timeline is the test case for whether principled AI development can survive political retaliation and commercial pressure. The February 2026 Pentagon standoff — when [[Pete Hegseth]]'s [[Department of Defense]] demanded [[Anthropic]] remove safety restrictions and [[Dario Amodei]] refused — is the single most consequential act of corporate resistance to the Trump administration in the AI sector. Sister timelines: [[OpenAI Timeline]], [[Defense Tech Timeline]], [[Anti-Trump Resistance Timeline]]. --- ## Era 1 — The Safety Schism and Founding (2021) - **January 2021** — [[Anthropic]] founded by [[Dario Amodei]] (CEO, former VP of Research at [[OpenAI]]) and [[Daniela Amodei]] (President), along with five other ex-[[OpenAI]] employees including Tom Brown (GPT-3 lead), Chris Olah, Sam McCandlish, Jared Kaplan, and Jack Clark. The departure is the institutional schism that defines AI governance: the people who understood [[OpenAI]]'s safety mission best concluded that commercialization had made it impossible to pursue from inside. - **RWKB angle:** This is the founding rupture that the [[OpenAI Timeline]] chronicles from the other side. [[Dario Amodei]] was VP of Research — he saw the commercial incentives overriding safety considerations firsthand. The founding of [[Anthropic]] is the moment the AI industry forked into two governance philosophies: build fast and monetize (OpenAI/[[Microsoft]]) vs. build carefully with structural constraints ([[Anthropic]]). Every subsequent development — the Claude models, the [[Constitutional AI]] methodology, the Pentagon standoff — follows from this fork. - **January 2021** — [[Anthropic]] incorporated as a **Public Benefit Corporation** from day one. Unlike a standard C-corp, a PBC is legally required to consider the impact of its decisions on society, not just shareholders. This is the structural choice [[OpenAI]] was supposed to make with its nonprofit governance — but [[OpenAI]]'s board was ousted when it tried to exercise that authority (see [[OpenAI Timeline]]). - **RWKB angle:** The PBC structure is philosophically significant but legally untested under extreme pressure. Whether a PBC charter actually constrains behavior when investors hold billions in equity is an open question — [[Anthropic]] is the experiment. - **May 2021** — [[Anthropic]] raises $124M seed round. Early backers include [[Google]], [[Effective Altruism]]-adjacent funders, and venture capital. The safety positioning attracts investors who want exposure to frontier AI without the governance risk of [[OpenAI]]. --- ## Era 2 — Constitutional AI and Building (2022) - **2022** — [[Anthropic]] develops [[Constitutional AI]] (CAI) — a training methodology where AI systems are aligned using a written set of principles rather than relying solely on human feedback. The AI evaluates its own outputs against the "constitution" and revises them. This approach is the methodological counterpart to the structural PBC commitment: embed values into the technical architecture, not just the corporate charter. - **RWKB angle:** [[Constitutional AI]] is a direct rejection of [[OpenAI]]'s approach of training AI through reinforcement learning from human feedback (RLHF) alone. The philosophical claim: you can encode ethical principles into AI the way you encode legal principles into a constitution. Whether a corporate-written AI constitution is more durable than [[OpenAI]]'s nonprofit charter is the deeper question — but at minimum, the technical approach provides a different failure mode than pure commercial optimization. - **2022** — [[Anthropic]] trains Claude 1 model internally — before ChatGPT launches in November. The lab is building frontier AI capabilities on par with [[OpenAI]] but without the same public profile. - **2022** — [[Anthropic]] raises $580M Series B. [[Google]] invests $300M, beginning what will become a multi-billion dollar strategic partnership. [[Google]]'s motivation: hedging against [[Microsoft]]'s exclusive partnership with [[OpenAI]]. [[Anthropic]] becomes the proxy battlefield for the AI arms race among tech giants — [[Amazon]] and [[Google]] compete to back [[Anthropic]] as a counterweight to [[Microsoft]]-[[OpenAI]] dominance. --- ## Era 3 — Claude Launch and Scaling (2023) - **March 2023** — **Claude 1 released publicly** — positioned as the safety-focused alternative to ChatGPT. The public launch comes four months after ChatGPT catalyzed the global AI arms race. [[Anthropic]] markets Claude's safety properties — refusal of harmful requests, reduced hallucination, transparent reasoning — as product differentiators rather than limitations. - **July 2023** — **Claude 2 released** — 100K token context window, improved reasoning capabilities. The context window is a significant technical advance, allowing Claude to process book-length documents in a single conversation. - **September 2023** — [[Amazon]] invests $4 billion in [[Anthropic]] — the first installment of what will eventually reach $8 billion. [[Amazon]] secures priority access to Claude models for [[Amazon]] Web Services customers. This mirrors the [[Microsoft]]-[[OpenAI]] structure: a cloud hyperscaler bankrolls a frontier AI lab in exchange for commercial distribution rights. - **RWKB angle:** The [[Amazon]] and [[Google]] investments create an interesting governance tension. [[Anthropic]] accepted billions from companies with their own [[Surveillance Capitalism]] business models and government contracting interests. Whether the PBC structure can resist pressure from investors who want unrestricted military and surveillance applications of Claude is the question that the 2026 Pentagon standoff answers. - **2023** — [[Google]] invests an additional $2 billion in [[Anthropic]], bringing total [[Google]] investment to over $2.3 billion. [[Anthropic]] is now funded by two of the three largest cloud providers ([[Google]] and [[Amazon]]), with only [[Microsoft]] (aligned with [[OpenAI]]) on the other side. - **November 2023** — **Claude 2.1 released** — 200K token context window, the largest commercially available at the time. [[Anthropic]] pushes the technical frontier while maintaining its safety positioning. --- ## Era 4 — Claude 3 and Commercial Growth (2024) - **March 2024** — **Claude 3 family launched**: Haiku (fast/cheap), Sonnet (balanced), and Opus (most capable). Claude 3 Opus matches or exceeds GPT-4 on multiple benchmarks, establishing [[Anthropic]] as a genuine peer competitor to [[OpenAI]] — not just a safety-focused niche player but a frontier lab producing the most capable models in the world. - **RWKB angle:** The Claude 3 launch demolishes the narrative that safety-focused development necessarily means inferior capability. [[Anthropic]] demonstrates that you can build frontier AI and maintain safety principles — the argument that safety and capability are zero-sum, used by [[OpenAI]] and others to justify dismantling safety teams, is empirically challenged. - **June 2024** — **Claude 3.5 Sonnet released** — widely adopted across the industry for strong coding performance and balanced capability. Becomes the default model for many developers and enterprises. - **2024** — [[Anthropic]] introduces the **Responsible Scaling Policy** (RSP) — voluntary commitments to test for catastrophic risks (biological weapons capability, cyber offense, autonomous replication) before deploying more powerful models. If testing reveals dangerous capabilities, [[Anthropic]] commits to not deploying until safety measures are adequate. - **RWKB angle:** [[Anthropic]] is the only major AI lab that voluntarily limits capability deployment based on safety testing. Whether this holds as competitive pressure mounts is the key question — and the February 2026 Pentagon standoff provides the answer: it held, at enormous cost. - **October 2024** — **Claude 3.5 Sonnet (upgraded) with computer use capability** — Claude can now operate a computer interface, clicking buttons, filling forms, and navigating applications. This represents a step toward AI agents that can act autonomously in digital environments. - **2024** — [[Amazon]]'s total investment in [[Anthropic]] reaches $8 billion across multiple installments, with [[Amazon]] securing up to 5 GW of compute capacity for [[Anthropic]]'s training needs. --- ## Era 5 — Massive Scaling (2025) - **March 2025** — [[Anthropic]] raises $3.5 billion at a $61.5 billion valuation. Revenue growth is explosive — [[Anthropic]] will achieve 10x annual revenue growth for three consecutive years. - **May 2025** — **Claude 4 released** — a major generational leap in reasoning, coding, and instruction-following. Claude 4 establishes [[Anthropic]] as an unambiguous frontier lab. - **September 2025** — $13 billion Series F at $183 billion valuation. [[Anthropic]]'s valuation trajectory begins approaching the hyperscalers themselves. - **October 2025** — [[Google]] partnership expansion — [[Anthropic]] secures access to up to 1 million TPUs (Tensor Processing Units), one of the largest single compute commitments in AI history. - **2025** — Revenue run rate reaches $14 billion — 10x annual growth sustained over three years. [[Anthropic]] transitions from well-funded startup to one of the most valuable private companies in the world. --- ## Era 6 — The Pentagon Standoff (2026) - **January 2026** — [[Anthropic]] introduces internal "Labs" division, reorganizing research and development operations. - **February 2026** — **THE DEFINING MOMENT.** [[Pete Hegseth]]'s [[Department of Defense]] demands [[Anthropic]] remove contractual bans on Claude being used for mass domestic surveillance and fully autonomous weapons systems. [[Dario Amodei]] refuses. This is the decision that triggers the most consequential confrontation between a tech company and the federal government in the AI era. - **RWKB angle:** Every other major tech company capitulated. [[Meta]]'s [[Mark Zuckerberg]], [[Amazon]]'s [[Jeff Bezos]], [[Google]]'s leadership, and [[OpenAI]]'s [[Sam Altman]] all donated to [[Donald Trump]]'s inauguration and accepted government terms. [[Anthropic]] refused to remove safety guardrails — and got blacklisted. The Pentagon standoff is the clearest illustration of how the Trump administration rewards compliance and punishes principled resistance in the [[Tech Oligarchy]]. - **February 2026** — [[Donald Trump]] directs all federal agencies to "immediately cease" using [[Anthropic]] technology. The [[Department of Defense]] designates [[Anthropic]] a **"supply chain risk"** — a label normally reserved for adversarial nations like [[China]] and Russia. The designation is an act of economic warfare against a domestic company for exercising its contractual rights. - **RWKB angle:** The "supply chain risk" designation is [[Authoritarianism]] in corporate-regulatory drag. The label exists to protect against adversary-nation infiltration of defense procurement. Using it against an American company because its CEO refused to allow autonomous weapons is the definition of political retaliation — and a federal judge agreed (see below). - **February 27–28, 2026** — [[OpenAI]] announces Pentagon classified network deal **the day after** [[Donald Trump]] bans [[Anthropic]]. The commercial reward for compliance is made explicit: refuse the government's demands and get blacklisted; accept them and get the contract. [[OpenAI]] accepted the "all lawful use" terms that [[Anthropic]] refused. See [[OpenAI Timeline]], [[Defense Tech Timeline]]. - **RWKB angle:** The timing is not coincidental. The Pentagon awarded the contract to [[OpenAI]] within 24 hours of banning [[Anthropic]] — a signal to every other AI company that compliance equals contracts and resistance equals economic punishment. This is the [[Military-Industrial Complex]] absorbing the AI industry in real time. - **February 2026** — Despite the federal ban, [[Anthropic]] raises $30 billion Series G at $380 billion valuation. Investors back [[Anthropic]] *because of* the government confrontation, not despite it — betting that principled AI development is commercially viable and that the ban will be struck down or reversed. - **RWKB angle:** The Series G is a market referendum on the Pentagon standoff. Capital markets are betting on [[Anthropic]]'s legal position and long-term commercial viability over the Trump administration's retaliatory power. This is significant: when [[Effective Altruism]]-adjacent institutions and major investors put $30 billion behind a company the US government just blacklisted, the government's economic coercion strategy has limits. - **March 2026** — [[Anthropic]] files **two federal lawsuits** against the Trump administration, alleging illegal retaliation for the company's safety stance. The suits challenge the supply chain risk designation and the contract ban as unconstitutional punishment for protected corporate speech. - **March 2026** — [[Anthropic]] launches the **Anthropic Institute**, a think tank focused on AI governance, led by Jack Clark (co-founder). The Institute positions [[Anthropic]] as not just a commercial competitor to [[OpenAI]] but an intellectual counterweight on AI policy — making the governance argument while the lawsuits make the legal argument. - **March 26, 2026** — A federal judge **temporarily blocks** the supply chain risk designation, calling it **"classic [[First Amendment]] retaliation."** The ruling holds that the government cannot use economic punishment to coerce a company into abandoning its safety policies. This creates potential legal precedent for corporate speech protections against government economic punishment — significant beyond AI. - **RWKB angle:** The [[First Amendment]] ruling is the legal system pushing back against [[Authoritarianism]]. The judge's language — "classic First Amendment retaliation" — frames the Pentagon standoff not as a procurement dispute but as a free speech case. If this precedent holds, it constrains the government's ability to use contract denial as punishment for corporate policy positions — a tool the Trump administration has deployed against multiple companies and institutions. - **May 2026** — [[Anthropic]] raises $65 billion Series H at **$965 billion valuation**. The company approaches a trillion-dollar valuation while actively suing the federal government. Revenue growth continues at approximately 10x annual pace. - **June 1, 2026** — [[Anthropic]] files **confidential IPO paperwork**. The company prepares to go public as one of the most valuable private companies in history — and one actively engaged in federal litigation against the sitting administration. - **June 19, 2026** — **Nobel laureate [[John Jumper]] leaves [[Google DeepMind]] for [[Anthropic]].** The [[AlphaFold]] co-creator and 2024 Chemistry laureate becomes the marquee hire of Anthropic's 2025–2026 talent run, aligning with its expansion into life sciences and computational biology. Paired with [[Noam Shazeer]]'s departure for [[OpenAI]] days earlier, the dual loss erased $200B+ from [[Alphabet]]'s market value in a week. - **RWKB angle:** That a sitting Nobel laureate would join a company *banned from federal contracts and suing the administration* is itself a market signal: elite research talent is betting on Anthropic's long-term position over the security of the world's dominant AI lab. The hire also opens a new competitive front — AI-for-science — where Anthropic now challenges [[Google DeepMind]] directly, and reframes the "compliant vs. principled" fork (see below) as also a talent-magnetism story, not just a contracts story. - **2026** — **Opus 4.6 launched** — the most capable Claude model to date, advancing [[Anthropic]]'s position at the frontier of AI capability alongside [[OpenAI]] and [[Google DeepMind]]. - **June 27, 2026** — **Mythos 5 limited release approved.** The administration permits release to 100+ U.S. institutions; Fable remains unrestricted. Anthropic had disabled access to both models to comply with the [[Anthropic Model Export Directive]], the Commerce Department's "cyber weapons" export classification that barred provision to any foreign national — including Anthropic's own non-U.S. employees. - **August 27, 2026** — **The standoff ends in court, and Anthropic wins.** Judge **Rita F. Lin** (N.D. Cal.) grants [[Anthropic]] **summary judgment** in a 59-page order on [[First Amendment]] retaliation, [[Fifth Amendment]] due-process and [[Administrative Procedure Act]] grounds. She **vacates the supply-chain-risk designation** and **[[Pete Hegseth]]'s order barring military contractors from any business with Anthropic**, enters a **permanent injunction**, and **denies** the government a seven-day stay. Lin finds the measures were meant to make a public example of Anthropic for its "arrogance," not to address any security threat. She **rejects** the claim that [[Donald Trump]]'s directive exceeded presidential authority. - **RWKB angle:** the vacated **boycott order** is the piece that reached beyond the parties, and its collapse hands [[Palantir]] a windfall it never litigated for — the [[Maven Smart System]] runs Claude, and the order had pointed Palantir toward a costly model swap inside its flagship Pentagon program. A court found the government's national-security framing pretextual in a suit brought by the company that refused a mass-surveillance carve-out; the practical beneficiary is the surveillance-adjacent contractor that depends on its model. See [[Anthropic Government Ban]]. - **September 25, 2026** — **The D.C. Circuit upholds the exclusion under a second statute.** In *Anthropic PBC v. U.S. Department of War*, a divided panel (Judges Gregory Katsas and Neomi Rao, with Judge Karen LeCraft Henderson dissenting) holds that the Pentagon lawfully excluded [[Anthropic]] under the Federal Acquisition Supply Chain Security Act, 41 U.S.C. § 4713. The majority finds that Anthropic's built-in restrictions on certain military uses can count as a statutory "supply chain risk," and rejects the [[First Amendment]] claim because the exclusion rested on Anthropic's "refusal to assent to a contract term," not on its viewpoint. Henderson reads the statute's terms ("sabotage," "surveil") as requiring intentional subversive conduct, not a vendor declining certain uses. Under Secretary [[Emil Michael]] posts that "the hammer of justice has smashed" Anthropic's arguments. Anthropic says it is "considering all options, including further review." - **RWKB angle:** the two rulings do not directly conflict. The August N.D. Cal. judgment addressed one statutory route, and the D.C. Circuit decided the other. The practical result is that the exclusion stands on the second route while the first was vacated. Whether a vendor's own usage policy can be treated as a supply-chain threat is now the open question for any en banc or Supreme Court review. --- ## Why This Matters — RWKB Significance [[Anthropic]] is the counter-narrative to every other tech company timeline in this vault. Seven threads make it essential KB infrastructure: **1. The safety schism.** [[Anthropic]] exists because [[OpenAI]]'s safety researchers concluded they couldn't build safe AI inside a commercializing organization. [[Dario Amodei]] was VP of Research — he saw the commercial incentives overriding safety firsthand. The founding story IS the AI governance debate. Compare with the [[OpenAI Timeline]], where the researchers who stayed watched their safety teams get dissolved. **2. PBC vs. profit.** [[Anthropic]] chose Public Benefit Corporation structure from day one — a legal commitment to balancing shareholder returns with societal impact. Whether this structure actually constrains behavior under investor pressure is an ongoing test, but as of 2026, the PBC charter appears to have provided real structural support for [[Dario Amodei]]'s refusal to comply with [[Department of Defense]] demands. The contrast with [[OpenAI]]'s nonprofit-to-for-profit conversion (see [[OpenAI Timeline]]) is the structural governance experiment of the AI era. **3. The Pentagon standoff as litmus test.** In February 2026, [[Pete Hegseth]]'s [[Department of Defense]] demanded [[Anthropic]] allow Claude for mass domestic surveillance and fully autonomous weapons. [[Anthropic]] refused. [[Donald Trump]] blacklisted them. [[OpenAI]] got the Pentagon contract instead. This is the clearest illustration of how the government rewards compliance and punishes principled resistance in the [[Tech Oligarchy]] — and the most consequential act of corporate resistance to the Trump administration in the AI sector. **4. Corporate resistance to Trump.** [[Anthropic]] is the only major AI company that has actively sued the Trump administration rather than donating to his inauguration. [[Meta]]'s [[Mark Zuckerberg]], [[Amazon]]'s [[Jeff Bezos]], and [[OpenAI]]'s [[Sam Altman]] all capitulated. The contrast is stark and belongs in the [[Anti-Trump Resistance Timeline]]. **5. The investor bet on ethics.** [[Anthropic]] raised $65 billion at $965 billion valuation AFTER being banned by the US government. Investors are betting that principled AI development is commercially viable — or at least that the ban won't survive judicial review. The Series G and Series H rounds are market referenda on whether [[Authoritarianism]] can economically coerce frontier companies into compliance. **6. [[First Amendment]] implications.** A federal judge called the supply chain risk designation "classic [[First Amendment]] retaliation." This creates legal precedent for corporate speech protections against government economic punishment — significant beyond AI, applicable to any government attempt to punish companies for policy positions. **7. The AI governance fork.** The AI industry has split into two camps: companies that comply with government demands for unrestricted military AI ([[OpenAI]], [[Meta]], [[xAI]]) and companies that maintain safety restrictions ([[Anthropic]]). This fork — mapped across the [[OpenAI Timeline]], [[Defense Tech Timeline]], and this timeline — will define AI governance for decades. The question is whether the safety camp can survive when the compliance camp gets rewarded with government contracts and political favor. --- ## See Also - [[Anthropic]] — Company profile - [[Dario Amodei]] - [[Daniela Amodei]] - [[Constitutional AI]] - [[AI Safety]] - [[OpenAI]] — Primary competitor; accepted Pentagon terms [[Anthropic]] refused - [[OpenAI Timeline]] — The other side of the safety schism - [[Sam Altman]] - [[Pete Hegseth]] - [[Donald Trump]] - [[Department of Defense]] - [[First Amendment]] - [[xAI]] — Also accepted Pentagon "all lawful use" terms - [[Google]] — Major investor - [[Amazon]] — Major investor ($8B total) - [[Google DeepMind]] — Competitor - [[Microsoft]] — Aligned with [[OpenAI]] - [[Meta]] — Competitor; capitulated to Trump - [[Stargate Project]] - [[AI Political Spending Ecosystem]] - [[Dual-Use AI]] - [[Defense Tech Conflicts of Interest]] - [[Military-Industrial Complex]] - [[Tech Oligarchy]] - [[Effective Altruism]] - [[TESCREAL]] - [[Regulatory Capture]] - [[Surveillance Capitalism]] - [[Authoritarianism]] - [[Defense Tech Timeline]] - [[Anti-Trump Resistance Timeline]] - [[Thiel Political Donations Timeline]] - [[Facebook Timeline]] - [[Twitter Timeline]] - [[Amazon Timeline]]