Who owns Claude? Origin, founders and the history of Anthropic

Who owns Claude? Anthropic, the American company that develops it, founded on January 26, 2021 in San Francisco, California. Its co-founders are former OpenAI staff, including Dario Amodei, its CEO, and his sister Daniela Amodei, its president. The first Claude model was released on March 14, 2023; the most recent, Claude Sonnet 5.5, dates from September 28, 2026.

I use Claude every day, and much of what circulates about its origins is approximate: nationality, creators, the meaning of the name, owners. Here are the facts as verified on October 7, 2026, with their official sources where they exist, and what remains a hypothesis.

Where is Claude AI from?

Claude AI is of American origin. Anthropic is headquartered at 500 Howard Street in San Francisco, and it is a Delaware company set up as a public benefit corporation. The headquarters and the leadership are in the United States.

Claude is a very common first name in France, which fuels confusion. The question "is Claude AI French?" even shows up among Google's related questions in French search results. The answer is no. The best-known assistant of French origin is Le Chat, from Mistral AI, a startup founded in Paris in 2023.

Anthropic does, however, have a presence in Europe. In November 2025, the company announced the opening of offices in Paris and Munich, in addition to London, Dublin and Zurich. Claude has been available in France since May 14, 2024, the date of its European launch on the web and on iPhone.

Anthropic, Claude, claude.ai: who is who?

These names are constantly mixed up, including in the press. Here is how they break down:

• Anthropic is the company.
• Claude refers to the family of AI models (Fable, Opus, Sonnet, Haiku) and, by extension, the assistant.
• claude.ai and the Claude apps are the conversational assistant, on the web, desktop and mobile. There is a free plan, whose limits I detail in Is Claude free?, and paid plans, compared in Claude pricing.
• Claude Code is the coding tool that puts Claude to work in a terminal or an editor. Getting started is explained in my guide to installing Claude Code.

Who made Claude? Anthropic's founders

Claude was created by Anthropic, founded by a group of former OpenAI employees. Eight co-founders are generally cited:

Several of them worked on GPT-2 and GPT-3. Anthropic was born when they left OpenAI, with a project centered on the safety of AI models. Claude and ChatGPT therefore share part of their history: I tell it from the OpenAI side in my article on who created ChatGPT.

A public benefit corporation, not a classic startup

Anthropic is set up as a public benefit corporation: a for-profit company whose charter requires it to pursue a public-interest goal as well, alongside shareholders' interests. On top of that sits the Long-Term Benefit Trust, an independent body tasked with appointing part of the board of directors.

These are legal constraints, not guarantees of behavior. They do, however, explain the company's messaging, heavily focused on safety, and some of its choices, such as the conditions it set for the Pentagon in 2026 (see below).

History of Anthropic: from its founding to 2026

Anthropic's history falls into four periods. The slider below walks through them step by step, and the button copies the full timeline.

2021-2022: research before product

For two years, Anthropic sold nothing to the general public. The team published research, including, at the end of 2022, the method known as "constitutional AI": the model is trained to follow a list of written principles, its "constitution". That constitution still exists: Anthropic published a much expanded version of it in 2026.

2023-2024: Claude arrives, then settles in Europe

Claude and Claude Instant were released on March 14, 2023, on request for access. Claude 2, on July 11, 2023, opened claude.ai to the public in the United States and the United Kingdom. In the fall of 2023, Amazon and then Google announced investments worth several billion dollars. On March 4, 2024, the Claude 3 family established the three names Opus, Sonnet and Haiku. On May 14, 2024, Claude arrived in Europe, and in June Claude 3.5 Sonnet outperformed the most powerful model of the previous generation.

2025: the turn toward agents

On February 24, 2025, Claude 3.7 Sonnet came out alongside the first version of Claude Code. On May 22, Claude 4 marked the shift from dialogue to agent: the model chains actions, reads files and calls tools over long sessions. Opus 4.1, Sonnet 4.5, Haiku 4.5 and Opus 4.5 followed between August and November. In November 2025, Microsoft and Nvidia in turn announced investments.

2026: hypergrowth and standoffs with Washington

2026 packs in more events than the previous five years combined:

• February: Opus 4.6 and Sonnet 4.6, then a $30 billion raise (Series G).
• March: the Pentagon labels Anthropic a "supply chain risk" after the company refuses to let Claude be used for fully autonomous weapons or for mass surveillance of the American population. On March 26, a federal judge in San Francisco suspends the measure.
• April: Claude Mythos Preview, reserved for cybersecurity partners (Project Glasswing), then Opus 4.7.
• May: Opus 4.8 and a $65 billion raise (Series H).
• June: a confidential IPO filing, the release of Fable 5 and Mythos 5, the suspension of both models from June 12 to 30 on the order of the US government, then Sonnet 5.
• July to September: Opus 5, Fable 5.1 and Mythos 5.1, Opus 5.5 and Sonnet 5.5.

Timeline of Claude models, from 2023 to today

The dates below come from Anthropic's announcements. The comments on performance reflect what the company stated, not my own measurements.

As of October 7, 2026, Anthropic's documentation recommends starting with Claude Opus 5.5 for most uses, and keeping Fable 5.1 for the most demanding reasoning and long agent tasks. Haiku 4.5 remains the fast, low-cost model.

Fable, Mythos, Opus, Sonnet, Haiku: what do the names mean?

• Opus, Sonnet, Haiku (since March 2024): three literary forms, from longest to shortest. Opus is the most capable of its generation, Sonnet offers the best balance of speed and cost, Haiku is the fastest and cheapest.
• Mythos (since April 2026): the most powerful models, reserved for vetted organizations in cybersecurity and the life sciences.
• Fable (since June 2026): the same model as Mythos, with additional safeguards, open to everyone via the API and paid plans.

The number indicates the generation. "Opus 5.5" therefore reads as "the Opus model of generation 5.5". Remember the logic rather than the numbers, which now change several times a quarter.

Why is Claude called Claude?

Anthropic has never published a statement explaining the choice of name. The most widely cited explanation is a tribute to Claude Shannon (1916-2001), the American mathematician and engineer regarded as the father of information theory. The New York Times presented it this way in February 2026, in an article on the history of Bell Labs, where Shannon worked.

The connection makes sense. His 1948 paper, "A Mathematical Theory of Communication", laid the foundations for measuring information and already modeled English text as a sequence of probable letters and words. It is the distant ancestor of what a language model does when it predicts the next word.

Other reasons circulate, with no official source either: a human first name, easy to pronounce in many languages, that stands apart from assistants with female names such as Siri or Alexa. I mention them as hypotheses. An article that presents one of them as an established fact has not checked.

Who owns Claude? Anthropic's main investors

Claude is owned by Anthropic, which is not listed on the stock market. Its main investors, according to public announcements:

At the time of the Series H, Anthropic reported annualized revenue of more than $47 billion. On June 1, 2026, the company confidentially submitted a draft IPO registration statement to the SEC, the US securities regulator, without setting a date or a price.

These investors are also suppliers. Amazon Web Services is Anthropic's main cloud partner, Google supplies it with TPU chips, and Claude is offered on the three major clouds (AWS, Google Cloud, Microsoft Azure) in addition to Anthropic's own API.

Claude is American: what it means for users in France and Europe

The question comes up in every meeting where I suggest these tools to a French company. Here are the verifiable points, without dramatizing or downplaying.

Who is responsible for your data?

For users in the European Economic Area, the data controller is Anthropic Ireland, Limited, in Dublin. The privacy policy in force since September 10, 2026 states that personal data is transferred to servers in the United States or other countries outside the European Economic Area, with standard contractual clauses as a safeguard.

Are your conversations used to train Claude?

Since the update to the consumer terms announced on August 28, 2025, Anthropic trains its models on conversations from Free, Pro and Max accounts if the corresponding setting is turned on. In that case, retention increases to five years for new conversations; otherwise, it stays at 30 days. The setting can be changed at any time in the privacy settings. Business plans (Team, Enterprise) and the API are not affected.

Can you keep data in Europe?

Not with Anthropic's API directly. Its documentation is clear: processing can be "global" or limited to the United States (with a 10% surcharge), and data at rest is stored in the United States, the only region available. To keep processing in Europe, the usual route is to use Claude through Amazon Bedrock or Google Cloud, which offer regional endpoints. It is one of the points I settle in writing before any AI automation project with a client.

Dependence on decisions made in Washington

2026 showed it twice. The conflict with the Pentagon first led to a directive asking federal agencies to stop using Claude, which the courts suspended. Then, on June 12, the US government placed Fable 5 and Mythos 5 under export controls: Anthropic had to switch off both models for all its customers, worldwide, until June 30. The other models stayed available. The practical lesson: a business process that relies on Claude should plan for a fallback model.

Claude, ChatGPT, Gemini, Le Chat: where do they come from?

Claude and ChatGPT share the same geographic origin and part of the same human history, but their technical and commercial choices have diverged. To choose between the two for everyday work, I compared Claude vs ChatGPT use case by use case, and my Claude AI review details what it does best.

Written by

Anthony Courtin

SEO, GEO & AI automation consultant

I help SMBs and independents with their visibility (Google, ChatGPT, Perplexity) and with the real-world use of AI at work. I work with companies to train their teams and rethink their processes. Everything I write here, I use with clients.

Follow me on LinkedIn My YouTube channel (in French)