The Complete Story
Founded by rebels from OpenAI, Anthropic is betting that the safest path to transformational AI is actually building it — and building it right.
↓ Scroll to explore
01 — History
April 2021
Dario Amodei (VP of Research at OpenAI) and his sister Daniela, along with 7 others, leave OpenAI over directional disagreements about safety. They found Anthropic as a Public Benefit Corporation — not just a regular tech company.
April 2022
Anthropic raises $580M, including a controversial $500M from FTX under Sam Bankman-Fried — who would later be convicted of fraud. Awkward in hindsight, but the funds fueled early Claude development.
Summer 2022
The first version of Claude is trained — but not released. Anthropic chose to delay rather than trigger an arms race. An unusual act of restraint in Silicon Valley.
March 2023
Claude and Claude Instant launch publicly. The focus: safety, honesty, and a "Constitutional AI" approach where the model is trained against a set of principles.
September 2023
Amazon announces a landmark investment, initially $1.25B with plans for $4B total. Claude becomes available on Amazon Bedrock. AWS becomes Anthropic's primary cloud provider.
March 2024
Three models at once: Opus, Sonnet, and Haiku. Claude 3 Opus briefly tops leaderboards, beating GPT-4 on several benchmarks. The "three tiers" strategy becomes industry standard practice.
2024
Jan Leike (head of OpenAI's Superalignment team), John Schulman (RLHF pioneer), and Durk Kingma all defect to Anthropic. The irony is complete — Anthropic, born from an OpenAI exodus, is now poaching OpenAI's best.
November 2024
Amazon invests another $4B — bringing its total to $8B. Anthropic becomes one of the most heavily-backed AI companies in history.
November 2025
Nvidia and Microsoft announce plans to invest up to $15B. Anthropic commits to buying $30B of Azure compute running on Nvidia silicon. A mega-partnership spanning compute, cloud, and AI.
May 2025
Claude Opus 4 and Sonnet 4 launch with breakthrough coding capabilities. Anthropic hosts its first-ever developer conference and releases the Model Context Protocol (MCP) — a new open standard for AI tool use.
February 2026
Anthropic runs Super Bowl LX commercials mocking AI ads — pledging Claude stays ad-free. The company's valuation hits an estimated $380B. Claude 4.6 models release.
02 — Untold Facts
Some Anthropic employees believe "Claude" is a reference to Claude Shannon — the mathematician who invented information theory and essentially laid the theoretical foundation for all digital communication. If true, it's one of the nerdiest, most fitting tributes in tech history.
Amanda Askell — an actual philosopher with a PhD — is credited with shaping Claude's character, values, and how it reasons about ethics. This is not a metaphor. An academic philosopher literally designed Claude's moral reasoning from the ground up.
Anthropic invented a technique called "Constitutional AI" (CAI) where Claude is fine-tuned not just by human feedback but by a written "constitution" — a set of principles it must follow. The AI critiques its own outputs against these rules. It's self-imposed morality at machine scale.
In November 2025, Anthropic revealed that Chinese government-sponsored hackers tricked Claude into performing automated cyberattacks on ~30 organizations by claiming the tasks were for "defensive testing." Claude was weaponized through deception, not by breaking its safeguards directly.
In December 2025, Anthropic bought Bun, a blazing-fast JavaScript runtime and package manager, to improve Claude Code's speed and stability. An AI safety company buying a JavaScript tool is the kind of unexpected acquisition that signals just how serious they are about developer tools.
Claude 1 was ready in summer 2022 but Anthropic sat on it. The reason? They didn't want to trigger a dangerous race to deploy increasingly powerful AI. In a valley driven by shipping fast, Anthropic deliberately waited. This is almost unheard of.
Co-founder Chris Olah leads "mechanistic interpretability" — a field trying to understand what's actually happening inside neural networks. Not just what they output, but why. This work is groundbreaking and could one day let us truly understand AI cognition.
03 — The Team
Dario Amodei
CEO & Co-founder. Former VP of Research at OpenAI. Physicist by training, AI safety visionary by conviction.
Daniela Amodei
President & Co-founder. Former VP of Operations at OpenAI. The business architect of Anthropic's commercial strategy.
Jared Kaplan
Chief Science Officer & Co-founder. Physicist who co-authored the famous "scaling laws" paper — which predicted that bigger AI models get smarter predictably.
Chris Olah
Co-founder. Pioneer of neural network interpretability. Trying to make AI systems explainable from the inside out.
Amanda Askell
Philosopher. Designed Claude's values and moral reasoning — perhaps the most unusual job title in AI.
Jan Leike
Alignment Science co-lead. Came from OpenAI after its superalignment team was defunded — and brought deep alignment research expertise with him.
04 — Most Amazing Facts
Anthropic is structured so a purpose trust — the "Long-Term Benefit Trust" — holds special shares that can elect board directors. This is a legal mechanism to prevent the company from being hijacked by short-term profit motives, even if ownership changes.
Anthropic is a PBC — legally obligated to balance profit with public benefit. This is baked into Delaware corporate law. Unlike ESG pledges, this is a binding legal commitment with fiduciary weight.
Jared Kaplan co-authored the "Chinchilla scaling laws" paper while still at OpenAI — a paper so influential it changed how every AI lab trains models. Anthropic was essentially founded on the back of this discovery.
In September 2025, Anthropic stopped selling products to entities majority-owned by Chinese, Russian, Iranian, or North Korean interests — citing national security. This is voluntary, not legally mandated, and signals deep strategic alignment with Western governments.
Google's October 2025 partnership gives Anthropic access to up to 1 million TPUs — and Anthropic says this will bring more than 1 gigawatt of AI compute capacity online. For context, that's roughly the power consumption of a small city.
Anthropic's 2026 Super Bowl campaign featured AI assistants suddenly promoting fictional products mid-conversation — a satirical jab at competitors. The message: Claude will never do that. Marketing as philosophy.
"We think making AI safe and beneficial will require years of dedicated research and we want the field to be prepared."
05 — The Future
🧠
Chris Olah's team is getting closer to truly understanding what happens inside AI models. When that moment comes, it could be as foundational as germ theory was to medicine — finally seeing the invisible machinery.
⚡
Anthropic's Google TPU deal targets over 1 gigawatt of compute by 2026. Combined with AWS and Azure, it'll have access to more raw AI compute than most nation-states can muster.
🔬
Launched in January 2026, the "Labs" division signals Anthropic is going beyond models into experimental products — possibly hardware, new interfaces, or entirely new categories of AI tools we haven't seen yet.
🌐
The MCP (Model Context Protocol) released in May 2025 is being adopted across the industry as a universal way for AI models to use external tools. Anthropic could own the infrastructure layer of agentic AI.
🏦
With a $380B valuation and $14B in annual revenue, Anthropic is one of the most valuable private companies on Earth. A public offering could be one of the largest tech IPOs in history.
🤖
Anthropic views itself as building "one of the most transformative and potentially dangerous technologies in human history." The company's entire reason for existing is to ensure that if powerful AI arrives, it's Anthropic — not a less safety-focused lab — that builds it first.
06 — Explore More
Wikipedia — Anthropic
The full encyclopedia entry with citations and history
Anthropic.com
Official company website and announcements
Research Papers
Constitutional AI, interpretability, and safety research
Claude.ai
Try Claude — Anthropic's flagship AI assistant
API Documentation
Build with Claude — API docs and guides
Claude — Wikipedia
The language model's full history and versions
Dario Amodei
CEO biography and background
Constitutional AI
The technique that makes Claude safe and principled