What is Claude AI? A guide for European business

Someone in the Monday meeting says they now run their reports through Claude AI, and three colleagues nod while hoping nobody asks them what it is. Eurostat counted 20 per cent of EU enterprises with ten or more staff using AI in 2025, up from 13.5 per cent a year earlier, so that nod happens in a lot of rooms.¹ This guide answers what is Claude AI in plain words, who owns it and which models exist in 2026. It also covers what Claude costs in euros and where a European business has to be careful before anyone pastes in a customer file.

What is Claude AI? The short answer

Claude AI is Anthropic's AI assistant, a family of language models, meaning programs trained on huge amounts of text to read and write, plus the apps and tools built around them.² Anthropic's models overview lists four current models on 9 November 2026: Claude Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5.³ Each has a context window, meaning how much text it holds in view at once, of 1 million tokens. The web app lives at claude.ai, and the desktop and mobile apps open the same account.

A language model counts as a chatbot only in the narrow sense, because the same Claude also writes software in Claude Code and works through your files in Claude Cowork. Claude does far more than chat, although it is still no colleague who carries the responsibility for a result, so a person signs off every answer that leaves the building.

Who owns Claude AI? The company behind Claude

Anthropic owns Claude AI, a company with its headquarters in San Francisco that the siblings Dario and Daniela Amodei founded in 2021 together with several former OpenAI staff.⁴ Its legal form is a public benefit corporation, a company form that writes a public purpose into its charter, and its stated mission is the responsible development of advanced AI for the long-term benefit of humanity.⁵

Founders, staff and investors own the private company, and no public source we have read shows a single investor in control of it. Amazon has invested US$8 billion in Anthropic, which calls Amazon a minority investor in its own announcement.⁶ Google plans to invest up to US$40 billion according to Bloomberg, with US$30 billion of that tied to performance targets.⁷ In May 2026 Anthropic raised US$65 billion at a valuation of US$965 billion, a round that included US$15 billion pledged earlier by cloud providers.⁸ A Long-Term Benefit Trust, whose trustees hold no financial stake in the company, is set to elect a growing share of the board.⁹ On 18 October 2026 Anthropic also announced that 5 per cent of its shares will go to a Public Benefit Fund for public-interest work, subject to stockholder approval.¹⁰

For a European buyer the more useful facts sit closer to home, because Anthropic runs offices in London, Dublin, Zurich, Paris, Munich and Milan.¹¹¹² Customers in the European Economic Area, Switzerland and the UK sign their contract with Anthropic Ireland, Limited, and the Standard Contractual Clauses in its DPA run under Irish law.¹³¹⁴ Readers who want the European picture in full, from supported countries to invoices, can read Claude in Europe.

Is Claude safe enough for a European business?

The answer depends on the plan you buy and on what your staff paste in, so check five points before anyone puts a customer file into Claude.

Contract. The DPA, the data processing agreement that Article 28 of the GDPR asks for, comes with the Commercial Terms for Team, Enterprise and the API.¹³¹⁴ Your own systems use the API to call Claude, for example a support tool that drafts replies. The DPA includes the EU Standard Contractual Clauses, while Free, Pro and Max sit under the consumer terms that carry no processor contract at all.¹⁵

Training. Anthropic does not train its models on inputs and outputs from Team, Enterprise and the API by default.¹⁶ Free, Pro and Max carry an opt-out, a setting where you say no to training, so check that setting on every private account that staff use for work.¹⁷

Certificates. Anthropic lists SOC 2 Type I and II, ISO 27001:2022 and ISO/IEC 42001:2023, the standard for how an organisation governs its AI, for its commercial products, while the consumer plans fall outside that scope.¹⁸

Where the data sits. Anthropic stores data in the US, and its own API offers only global or US-only inference, so no Anthropic-hosted plan processes your prompts inside the EU.¹⁹ Teams that need EU processing call Claude through Amazon Bedrock's EU cross-region inference, which keeps each request inside six EU regions when called from an EU region.²⁰ The other route is the EU multi-region endpoint on Google Cloud, which carries the 5.5 models.²¹

Retention. Zero data retention, meaning the vendor keeps no prompts after processing unless the law requires it, is no default setting. Anthropic switches it on per organisation, and it does not cover chat on claude.ai or Cowork.²² Fable 5.2 keeps prompts and completions for at least 30 days by default, so work with personal data is safer on the 5.5 models.²³

Picture an HR manager in Lyon who pastes a written warning into Claude on a Thursday afternoon. On a private Pro account no processor contract covers that letter, while on the company's Team account the DPA does, and the HR manager still removes the employee's name whenever the task allows it.

Rule of thumb: personal data belongs in Team, Enterprise or the API and never in private accounts, which is the one sentence worth putting into your AI policy this week. The full GDPR walk-through sits in is Claude GDPR compliant, the contract itself in Anthropic's DPA explained and the location question in Claude EU data residency.

Which Claude models are there in 2026?

All four current models hold 1 million tokens in view at once, word pieces that add up to roughly 555,000 words.³ The table gives Anthropic's API price in US dollars per million tokens in and out, as listed on the day we checked.

Claude Fable 5.1

Fable 5.2, released on 9 November 2026, handles the heaviest jobs: demanding reasoning and long agent tasks, where the AI takes many steps on its own before it reports back. Anthropic recommends starting with Opus 5.5 and moving to Fable 5.2 only when Opus at higher effort, meaning more thinking time per answer, still falls short.³ Its sister model Mythos 5.2 is open only to verified organisations, so most European teams will never see it in their model menu.²⁴

Claude Opus 5.5

Opus 5.5 is the model Anthropic tells most teams to start with, released on 22 September 2026, and it always thinks before it answers, a mode that cannot be switched off.³²⁵ For a team that wants one model for contracts, analysis and drafts, Opus 5.5 is the default choice, and the price per token sits at half of Fable 5.2.

Claude Sonnet 5.5

Sonnet 5.5 is the faster and cheaper partner to Opus 5.5 at half the price per token, and it also runs on Amazon Bedrock, Google Cloud and Microsoft Foundry.³²⁶ The older Sonnet 4.5 retires on 30 November 2026, and Anthropic names Sonnet 5.5 as its replacement, so any internal tool still pinned to the old model needs a change this month.²⁷

Claude Haiku 5.5

Haiku 5.5 is Anthropic's fastest and cheapest model, built for classification, extraction and routing, meaning sorting tickets, pulling out fields and sending each case to the right queue.²⁸ The low price suits a thousand support emails that need a label each, while an annual report where every figure has to be right belongs on Opus.

Claude Code and Claude Cowork

Claude Code is Claude for software work: it runs in the terminal, in VS Code and in JetBrains, and it comes with Pro, Max, Team and Enterprise but not with Free.²⁹³⁰ The desktop sibling, Claude Cowork, is Claude for everyday office work, a mode that works through the files in a folder on your computer. Cowork needs a paid plan, and an admin can switch it off for the whole organisation.³¹ Our Claude Code guide and the Cowork setup guide each cover the first thirty minutes with the tool. Readers who never write code can start with Claude Code explained for non-developers, which shows what a manager or analyst builds with it. Anthropic also ships Claude Design for decks and prototypes and add-ins for Excel, Word, PowerPoint and Outlook, explained in what Claude Design is and Claude for Microsoft 365. The whole line-up with the plan each tool comes with sits in our map of the Claude product suite.

How does Claude AI work in practice?

Picture a bid manager in Rotterdam with an 80-page tender document due on Monday morning. The Friday evening used to go on hunting for requirements, and now they ask Claude to list every requirement with its page number and check the five that carry penalties against the original. Claude reads, the person judges, and the person still owns the bid when it goes out.

Choosing a model. Think of an office where Haiku sorts the post for one fortieth of the Opus price and Sonnet takes routine work for half of it. Opus takes whatever has to be right the first time, while Fable 5.2 is the specialist at twice the Opus price with 30 days of retention. Run a cheaper model on ten real examples before you pay for a bigger one, because the cheap one often passes.

Where Claude is weaker. Claude can be wrong like every language model, so a person checks figures, names and legal references before anything goes to a customer. Nothing about your company reaches Claude until you hand it material, and its reliable knowledge stops in June 2026.³ Give it your own documents, for example a knowledge base built from your PDFs, and ask it to name the page each answer comes from.

Languages. On Anthropic's Global-MMLU test Opus 5.5 scores 99.1 per cent of its English result in German and 99.3 per cent in French and Spanish.³² A team in Munich, Lyon or Madrid can therefore work in its own language with almost no loss against English. Dutch, Polish and Swedish are not in that table, which means you should test ten real texts in your language before a Claude draft goes to a customer.

What does Claude AI cost?

Claude AI costs €0 on Free, €18 a month on Pro and €18 per seat a month on Team with annual billing, before VAT as claude.com shows European visitors.³⁰ Team needs at least two seats, while Enterprise costs US$20 per seat a month plus usage with a minimum of 20 seats. Max at €90 or Ultra at €180 a month suits one person who works in Claude all day. Every plan, the VAT rules and worked totals for a team sit in our guide to Claude prices in 2026. What a private buyer pays per country and how a business adds its VAT ID for the reverse charge is in Claude and VAT in the EU.

Through satori-claude we set Claude up for you for €490 once and €14 per user a month, all excluding VAT. The licence comes on top at €22 for a Standard seat or €110 for Premium, on one invoice in euros. Our managed Claude licences page runs those numbers for a team of ten next to the price of buying direct.

Is Claude better than ChatGPT?

The answer depends on the job, so run the same three real tasks from last week through both tools before you buy seats for a whole team. Our team has spent more than 1,000 hours with ChatGPT and Claude at customers since 2024. Claude wins on long documents, analysis, code and sensitive data in that time, while ChatGPT stays ahead when a team mostly needs images in the same chat window. Teams that already pay for Microsoft 365 can weigh the third option in our Claude vs Copilot comparison, since Copilot's Cowork agent runs on Claude models.

Let the person who will use the result make the pick, because a winner on a benchmark means little when the tool sits unopened after week two. The short version is in is Claude better than ChatGPT, and the ChatGPT side of the market is covered in ChatGPT in Europe.

What does Claude AI mean for a European business?

Claude gives every colleague one extra reader for the price of a seat, as long as a person hands it the material and checks the answer before it goes anywhere.

Picture a company where everyone got a Claude account on a Monday and three months later nobody can say which task Claude has taken over, which is the rollout we see go wrong most often. Our advice runs the other way: pick one task that comes back every week, time it today and let three people run it with Claude for two weeks before you buy a single extra seat.

Teams that want help with that first step can book Claude team training from €280 per user for two hours with at least eight people, while single learners join the online Claude course for €640. Agents that finish whole tasks come next in AI agents for business, and a Claude consultant starts in the same place as this guide: with one task, not with a tool.

How do you get started with Claude AI?

Three routes lead in, and each one fits a different starting point. Sign up yourself on claude.ai, with Pro for one person and Team for a group of two or more. Hand the setup to us with satori-claude, where we set up the workspace, the licences and the admin for you. Start the whole team at once with satori-launch, our programme from €560 per user with at least eight users.

Write a made-up customer email with three questions, paste it into claude.ai and ask Claude to summarise what the customer wants answered. If all three questions come back you have found a first task to hand over, and if one is missing you know exactly what a person still needs to check.

Sources

Footnotes

1. Eurostat (2025). 20% of EU enterprises use AI technologies. https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20251211-2 ↩

2. Anthropic (2026). Claude product overview. https://claude.com/product/overview ↩

3. Anthropic (2026). Models overview. Claude Platform Docs, checked on 9 November 2026. https://platform.claude.com/docs/en/models/overview ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

4. Wikipedia (2026). Anthropic. https://en.wikipedia.org/wiki/Anthropic ↩

5. Anthropic (2026). Company. https://www.anthropic.com/company ↩

6. Anthropic (2024). Anthropic and Amazon expand compute partnership. https://www.anthropic.com/news/anthropic-amazon-compute ↩

7. Bloomberg (2026). Google plans to invest up to $40 billion in Anthropic. https://www.bloomberg.com/news/articles/2026-04-24/google-plans-to-invest-up-to-40-billion-in-anthropic ↩

8. Anthropic (2026). Series H, 28 May 2026. https://www.anthropic.com/news/series-h ↩

9. Anthropic (2023). The Long-Term Benefit Trust. https://www.anthropic.com/news/the-long-term-benefit-trust ↩

10. Anthropic (2026). Anthropic Public Benefit Fund, 18 October 2026. https://www.anthropic.com/news/public-benefit-fund ↩

11. Anthropic (2025). New offices in Paris and Munich expand Anthropic's European presence. https://www.anthropic.com/news/new-offices-in-paris-and-munich-expand-european-presence ↩

12. Anthropic (2026). Anthropic opens Milan office. https://www.anthropic.com/news/milan-office-opening ↩

13. Anthropic (2025). Commercial Terms of Service. https://www.anthropic.com/legal/commercial-terms ↩ ↩²

14. Anthropic (2025). Data Processing Addendum. https://www.anthropic.com/legal/data-processing-addendum ↩ ↩²

15. Anthropic (2026). How do I view and sign your Data Processing Addendum (DPA)? Anthropic Privacy Center. https://privacy.claude.com/en/articles/7996862-how-do-i-view-and-sign-your-data-processing-addendum-dpa ↩

16. Anthropic (2026). Is my data used for model training? (commercial products). Anthropic Privacy Center. https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training ↩

17. Anthropic (2026). Is my data used for model training? (consumer products). Anthropic Privacy Center. https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training ↩

18. Anthropic (2026). What certifications has Anthropic obtained? Anthropic Privacy Center. https://privacy.claude.com/en/articles/10015870-what-certifications-has-anthropic-obtained ↩

19. Anthropic (2026). Data residency. Claude Platform Docs. https://platform.claude.com/docs/en/manage-claude/data-residency ↩

20. Amazon Web Services (2026). Geographic cross-Region inference. Amazon Bedrock User Guide. https://docs.aws.amazon.com/bedrock/latest/userguide/geographic-cross-region-inference.html ↩

21. Google Cloud (2026). Data residency. Gemini Enterprise Agent Platform documentation. https://docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency ↩

22. Anthropic (2026). Zero data retention. Claude Code Docs. https://code.claude.com/docs/en/zero-data-retention ↩

23. Anthropic (2026). Covered Models. Claude Help Center. https://support.claude.com/en/articles/15425695 ↩

24. Anthropic (2026). Claude Mythos 5.2 overview. Claude Platform Docs. https://platform.claude.com/docs/en/models/mythos-5-2/overview ↩

25. Anthropic (2026). Introducing Claude Opus 5.5. https://www.anthropic.com/claude-opus-5-5 ↩

26. Anthropic (2026). Introducing Claude Sonnet 5.5. https://www.anthropic.com/news/claude-sonnet-5-5 ↩

27. Anthropic (2026). Model deprecations. Claude Platform Docs, checked on 9 November 2026. https://platform.claude.com/docs/en/about-claude/model-deprecations ↩

28. Anthropic (2026). Claude Haiku 5.5 overview. Claude Platform Docs. https://platform.claude.com/docs/en/models/haiku-5-5/overview ↩

29. Anthropic (2026). Claude Code. https://claude.com/product/claude-code ↩

30. Anthropic (2026). Plans and Pricing, checked from an EU connection on 9 November 2026. https://claude.com/pricing ↩ ↩²

31. Anthropic (2026). Claude Cowork. https://claude.com/product/cowork ↩

32. Anthropic (2026). Multilingual support. Claude Platform Docs, checked on 9 November 2026. https://platform.claude.com/docs/en/build-with-claude/multilingual-support ↩