Claude for Enterprise: Anthropic, GDPR, DPA and EU Hosting (2026)

Claude is the AI model family of the US company Anthropic. Enterprises use Claude as a chat assistant, through the API in their own applications and with the coding agent Claude Code. Whether you are a corporation, a mid-sized company or a small business, the same questions come up before rollout. Who is behind Claude? Is there a data processing agreement (DPA)? Does Claude have servers in the EU? Which licence is the right one, and can Claude be bought on invoice? This guide answers them based on primary sources from Anthropic, AWS and Google Cloud.

Key facts (as of October 2026)

• Claude is developed by Anthropic, PBC, a public benefit corporation founded in 2021 and headquartered in San Francisco; according to the Commercial Terms, the contracting entity for customers in the EEA, Switzerland and the United Kingdom is Anthropic Ireland, Limited in Dublin.
• Claude can be used in a GDPR-compliant way if the access route and the contract are right: Claude Team, Claude Enterprise and the Claude API are covered by Anthropic’s Data Processing Addendum with EU standard contractual clauses, and according to Anthropic customer content is not used for training; the consumer plans Free, Pro and Max are not suitable for business data.
• Anthropic’s own API and claude.ai offer no EU data residency, and according to Anthropic data from claude.ai, Team and Enterprise is stored in the US; processing in the EU is available through AWS Bedrock (EU geo inference profile) and through the Gemini Enterprise Agent Platform on Google Cloud (EU multi-region).
• For employees’ chat workplace, companies license Claude Team or Claude Enterprise directly from Anthropic; for their own applications and EU hosting, they use the Claude models via API through AWS Bedrock or Google Cloud.
• In Germany, Claude can be paid for through the cloud invoice: innFactory GmbH is a cloud reseller for AWS, Microsoft Azure, Google Cloud and STACKIT and bills by invoice or direct debit, the innFactory AI Gateway adds budgets, SSO and cost control, and CompanyGPT provides the chat interface in your own tenant.

At innFactory AI we run Claude in production customer projects — as part of multi-LLM platforms like CompanyGPT and for coding agents via the AI Gateway. This article gives the practical view: what works, where the limits are and how to set Claude up correctly.

Which company is behind Claude?

Claude is developed by Anthropic, PBC. According to the company, Anthropic is a public benefit corporation (PBC) headquartered in San Francisco, California, and was founded in early 2021. The co-founders are Dario Amodei (CEO), Daniela Amodei (President and Chair of the Board), Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Chris Olah.

Governance: As a public benefit corporation, Anthropic says it is structured to balance commercial growth with a public benefit purpose. The Long-Term Benefit Trust, an independent body, has the authority to appoint a majority of Anthropic’s board, according to the company. Trustees include Neil Buddy Shah (Chair), Richard Fontaine, Ben Bernanke and, since 21 October 2026, the physicist Christopher Stubbs. On 18 October 2026 Anthropic also announced the Anthropic Public Benefit Fund, to which 5% of the company’s shares are to be committed, subject to stockholder approval.

Investors: In its Series H on 28 May 2026, Anthropic says it raised USD 65 billion at a post-money valuation of USD 965 billion. This includes USD 15 billion of previously committed investments from hyperscalers, of which USD 5 billion came from Amazon. According to Anthropic, AWS is its primary cloud and training partner; Google already took part in the Series C (2023).

Anthropic in Europe: Under the Commercial Terms, companies based in the EEA, Switzerland or the United Kingdom contract with Anthropic Ireland, Limited, and Irish law applies. For the consumer plans in the EEA and Switzerland, the contracting entity is also Anthropic Ireland, Limited. Where Anthropic itself acts as controller, for example for the consumer plans, the privacy policy (effective 10 September 2026) names Anthropic Ireland, Limited in Dublin as controller for users in the EEA, UK and Switzerland. According to the Trust Center, the group includes Anthropic Limited (UK), Anthropic Switzerland GmbH, Anthropic France SAS and Anthropic Germany GmbH, among others. In addition to London, Dublin and Zurich, Anthropic announced offices in Paris and Munich in November 2025 and, in May 2026, Milan as its sixth European location. Germany, Austria and Switzerland are officially supported regions. Relevant for the EU AI Act: Anthropic Ireland Limited has signed the General-Purpose AI Code of Practice.

Important for your data protection assessment: a German subsidiary and an office in Munich do not mean that data is processed in Germany. Where Claude processes your data depends solely on the access route (see the section “Where are Claude’s servers?” below).

Claude licences for business: Team, Enterprise and API

Anthropic distinguishes between consumer plans (Free, Pro, Max) for individuals and commercial offerings for organisations. The business plans for the chat workplace are Claude Team and Claude Enterprise; for your own applications there is the Claude API (Claude Platform). Only the commercial offerings are governed by the Commercial Terms and the data processing agreement. Anthropic publishes the prices for Team and Enterprise on its own website (see provider).

Source: Anthropic Privacy Center and Claude Platform documentation, as of 23 October 2026.

According to Anthropic, Claude Enterprise additionally offers custom retention periods, the Compliance API for connecting security and compliance tools, role-based access controls with groups from your identity provider (SCIM), HIPAA-ready configurations and, for usage-based plans, the “US-only inference” setting, among other things.

Claude API: inputs and outputs are deleted from the backend within 30 days; exceptions include the Files API, Usage Policy violations (up to 2 years) and legal obligations. Anthropic grants zero data retention (ZDR) only on request per organisation, for the Messages API, token counting and Claude Code with commercial API keys. According to Anthropic, not covered are the Console and Playground, Managed Agents, code execution and the Batch API, among others. The Team and Enterprise chat interfaces are not ZDR-eligible; the exception is Claude Code on Enterprise plans.

Covered models: for Claude Fable 5.1, Mythos 5.1, Fable 5 and Mythos 5, Anthropic generally requires 30-day retention of prompts and outputs on every access route.

Claude Enterprise is therefore a standalone offering for the chat workplace and at the same time an alternative to platforms such as CompanyGPT. Which option fits depends mainly on three questions: must processing take place in the EU? Should other models such as GPT, Gemini or open models be available alongside Claude? And should chat data reside in your own cloud tenant? A side-by-side comparison of both approaches is available at CompanyGPT vs. Claude.

Is Claude GDPR-compliant? DPA, standard contractual clauses and training

Claude can be used in a GDPR-compliant way, but not automatically. Three things matter: a commercial contract with a data processing agreement (DPA), control over training and retention, and a properly documented data flow, either with processing in the EU or with a safeguarded third-country transfer.

The DPA with Anthropic (Data Processing Addendum)

Anthropic calls its DPA the Data Processing Addendum. According to Anthropic, it is incorporated into the Commercial Terms and therefore applies to Claude Team, Claude Enterprise and the Claude API. The key provisions of the version effective 24 February 2025:

• Roles: the customer is the controller and Anthropic its processor (“Customer is the controller and Anthropic is Customer’s processor”).
• Standard contractual clauses: Module Two (controller to processor) and Module Three (processor to processor) of the EU standard contractual clauses are deemed concluded, supplemented by addenda for the UK and Switzerland.
• Security breaches: Anthropic notifies customers without undue delay, and in any event within 48 hours.
• Sub-processors: customers can object to new sub-processors within 15 days. According to the Trust Center, infrastructure providers include Google Cloud, AWS and Microsoft Azure (worldwide) as well as Cloudflare, plus further US service providers.
• Evidence: Anthropic supports transfer impact assessments and provides its SOC 2 report on request. According to the Trust Center, it holds SOC 2 Type 2, ISO 27001, ISO 42001, CSA STAR and NIST 800-171; for HIPAA there is a BAA for the API and Enterprise.
• Third-country transfers: the privacy policy and the DPA rely on adequacy decisions and standard contractual clauses. As with any US provider, the assessment of US access rights such as the CLOUD Act belongs in your transfer impact assessment.

Is business data used for training?

Not with the commercial offerings. The Commercial Terms state: “Anthropic may not train models on Customer Content from Services.” The exception is feedback that users actively submit via the thumbs button: Anthropic stores the related conversation for up to five years and may use it for training. Owners of Team and Enterprise plans can disable this feature in the organisation settings.

Why Free, Pro and Max are not suitable for business data

On the consumer plans, users have decided for themselves since 28 August 2025 whether their chats are used for training. If training is allowed, Anthropic stores the data in de-identified form for up to five years; otherwise deleted chats are removed within 30 days, and incognito chats are never used for training. The Consumer Terms apply instead of the Commercial Terms, and Anthropic is the controller itself. The company therefore has no central control over training and storage, which is not viable for personal or confidential business data.

Practical note: who do you sign the DPA with?

The contracting party depends on the access route. With Claude Team, Claude Enterprise, the Claude API, Claude Platform on AWS and Claude Models on Google Cloud (Anthropic managed), Anthropic is your processor and Anthropic’s DPA applies. If you use Claude through AWS Bedrock or the Gemini Enterprise Agent Platform (Google managed), you sign the DPA with the cloud provider, i.e. AWS or Google Cloud.

Where are Claude’s servers? EU hosting at a glance

According to Anthropic, claude.ai routes traffic from commercial customers by default to select countries in the US, Europe, Asia and Australia, with the note “Note that data is stored in the US.” As a control, Anthropic offers “US-only inference” for usage-based Enterprise plans; there is no EU option. Anyone who needs processing in the EU therefore uses a cloud provider.

Sources: Claude Platform documentation, Anthropic Privacy Center, AWS Bedrock model cards, Google Cloud data residency table, Microsoft Learn; as of 23 October 2026.

Five details that are often misrepresented in practice:

• Frankfurt is not an in-region location for Claude. On AWS Bedrock, Frankfurt (eu-central-1) is not a fixed processing location for any Claude model, only a source region of the EU geo profile.
• London is not in the EU. Since 30 September 2026, AWS has offered in-region processing in London (eu-west-2) for Opus 5.5 and Sonnet 5 (“processing never leaves that Region”). London is in the United Kingdom, i.e. outside the EU.
• Google’s EU multi-region covers neither the UK nor Switzerland.
• Google now has two Claude routes. “Google managed” runs on the Gemini Enterprise Agent Platform and supports the EU multi-region; “Anthropic managed” (since 22 October 2026) is Anthropic’s Claude API and offers no EU data residency. Check which variant you have booked.
• Bedrock stores prompts and outputs only for Fable 5 and 5.1 (30 days). For all other Claude models, AWS states “zero operator access”; model providers have no access.

We track which Claude version is available in the EU on which route, model by model, on the Anthropic Claude page.

How do enterprises use Claude in a GDPR-compliant way? Three routes

Direct access via claude.ai or the Anthropic API is possible with the DPA and standard contractual clauses, but offers no EU data residency. For business data with EU requirements we recommend one of the following three routes.

Route 1: AWS Bedrock with the EU geo inference profile

On AWS Bedrock, the current models Opus 5.5, Sonnet 5.5 and Haiku 5.5 run in the EU via the EU geo inference profile (model IDs with the eu. prefix) on the standard bedrock-runtime endpoint, i.e. with the Converse API and Guardrails. Routing stays within the EU geography. A fixed location in Ireland (eu-west-1) or Stockholm (eu-north-1) is only available via the bedrock-mantle endpoint (Anthropic Messages API) and only for Opus 5, Sonnet 5, Opus 4.8, Opus 4.7 and Haiku 4.5. Bedrock offers Fable 5.1 only with the global. and us. profiles. You sign the DPA with AWS.

Pros: established compliance, DPA with AWS, Converse API and Guardrails, “zero operator access” for all Claude models except Fable. Cons: no fixed EU location for the 5.5 models, in-region only with older models via bedrock-mantle, no EU profile for Fable 5.1; as a US provider, AWS is subject to US law, which needs to be considered in the transfer assessment.

Details on Claude on AWS and on billing by invoice are available from innFactory GmbH.

Route 2: Gemini Enterprise Agent Platform with the EU multi-region

At Google, the former Vertex AI route is now called “Claude Models on Google Cloud (Google managed)” and runs on the Gemini Enterprise Agent Platform. Opus 5.5, Sonnet 5.5, Haiku 5.5, Opus 5, Sonnet 5 as well as Fable 5 and 5.1 are available there in the EU multi-region, with ML processing in the EU; Google charges a 10% surcharge over the global endpoint. Fable 5.1 is therefore available in the EU via Google, whereas Bedrock only offers the global. and us. profiles. Google’s data residency table lists a single-region location in Belgium (europe-west1) for older models such as Haiku 4.5, Opus 4.5, Opus 4.6, Sonnet 4.5 and Sonnet 4.6; some model pages only mention the Europe multi-region, so check the specific model page before rollout. You sign the DPA with Google Cloud.

Watch out when booking: the variant “Claude Models on Google Cloud (Anthropic managed)”, available since 22 October 2026, is Anthropic’s Claude API reached through your Google Cloud project. Anthropic is the processor there, and there is no EU data residency.

Route 3: AI platform inside your own cloud tenant with integrated Claude access

Our CompanyGPT platform integrates Claude via Bedrock or the Gemini Enterprise Agent Platform directly inside the customer’s cloud tenant — chat and RAG data stay in the customer tenant, and the enterprise capabilities that plain model access does not provide (audit, RBAC, knowledge base, workflows, n8n automation) come out of the box. With this model (see our comparisons to Langdock, Logicc and neuland.ai) no tenant moves to a third party — tokens are billed at the hyperscaler’s list price, with no surcharge by innFactory. For many companies this is a fast and well-controlled route to Claude in production.

In CompanyGPT you can preconfigure the default model per use case: a fast model for routine tasks, Sonnet or Opus for complex work, plus models from OpenAI and Google in parallel. That way you only pay for the capability a task actually needs.

Buying Claude by invoice via innFactory

For many procurement teams, billing matters as much as data protection: invoice instead of credit card, a German contact and one invoice for all cloud services. Our parent company innFactory GmbH is a cloud reseller for AWS, Microsoft Azure, Google Cloud and STACKIT. It bills the cloud accounts in which Claude runs via AWS Bedrock or the Gemini Enterprise Agent Platform by invoice or direct debit — no credit card required, with all cloud providers on one consolidated invoice and a personal contact.

For control in day-to-day operation there is the innFactory AI Gateway. It runs as an infrastructure component in the customer’s tenant (not SaaS) and enforces budgets and hard spending caps per team and member, with single sign-on via Entra ID, Keycloak or Cognito. As providers it connects, among others, Anthropic Claude directly or via the Gemini Enterprise Agent Platform (formerly Vertex AI) and AWS Bedrock including Bedrock Mantle. Ready-made setups are available for Claude Code, Codex, OpenCode and Cowork (Claude Desktop). The Companion tracks subscription usage such as Claude Max separately and measures agent time per repository or project.

One limitation you should know: the classic chat window of Claude Desktop cannot be routed through a gateway. The agentic areas Cowork and Code can run through a gateway or a cloud provider (see Claude 3P with CompanyGPT); for chat in your own tenant there is CompanyGPT.

Which Claude models for enterprises?

The current family consists of Claude Opus 5.5 (since 22 September 2026), Claude Sonnet 5.5 (since 28 September 2026) and Claude Haiku 5.5 (since 7 October 2026). In addition, there is Claude Fable 5.1 (since 1 September 2026) from the Mythos class, which as a covered model is subject to mandatory 30-day retention. The 5.5 models have a context window of 1 million tokens.

Prices according to Anthropic for the Claude API, as of 23 October 2026. On Bedrock and the Agent Platform, the cloud providers’ price lists apply; Google’s EU multi-region costs a 10% surcharge, and the “us” inference geo at Anthropic costs 1.1×. Fast mode for Opus 5.5 has been generally available since 15 October 2026 (USD 8 / 40).

Our recommendation:

• Claude Sonnet 5.5 as the default for most knowledge-worker tasks
• Claude Opus 5.5 for deep reasoning, legal texts and complex engineering; our default for GDPR-sensitive workloads because it is not a covered model and zero data retention is possible
• Claude Haiku 5.5 for high-frequency, simple tasks
• Claude Fable 5.1 only selectively, because as a covered model 30-day retention applies on every route
• Claude Opus 5 or Sonnet 5 if a fixed location in Ireland or Stockholm on Bedrock is mandatory (via bedrock-mantle)

Mind the lifecycle: Claude Sonnet 4.5 has been deprecated since 30 September 2026 and will be retired on the Claude API on 30 November 2026 (replacement: Sonnet 5.5); on Bedrock the legacy version runs until 8 April 2027. Anthropic announces retirements at least 60 days in advance. All versions, benchmarks and EU details are on the Anthropic Claude model page.

Claude, GPT or Gemini: which model fits when?

Earlier versions of this article contained a rating table based on our own A/B tests in the first half of 2026 (Claude 4.x, GPT-5, Gemini 2.5). Since then, Anthropic, OpenAI and Google have released new model generations, and those results cannot be transferred to the current versions. We therefore no longer publish a blanket ranking.

For model selection in the enterprise, five criteria are more meaningful than benchmark positions:

1. EU availability of the specific model version on the access route you use
2. Retention and training, for example covered models or the availability of zero data retention
3. Cost per completed task rather than price per token
4. Integration with your existing environment (Microsoft 365, Google Workspace, your own systems)
5. Results with your own data: test two or three models with real tasks from your departments

We maintain current versions, prices and EU routes on the model pages for OpenAI GPT, Google Gemini and Anthropic Claude. How to choose models by department is covered in the article Claude or ChatGPT for Business?. In practice, the answer is rarely “only Claude” or “only GPT”: a multi-LLM strategy keeps your options open, even when a model is retired. More on this in our guide to AI for companies.

Typical use cases for Claude in the enterprise

Contract analysis and legal review

With a context of 1 million tokens, Opus 5.5 and Sonnet 5.5 can process extensive sets of contracts in a single request, such as framework agreements, arbitration clauses or shareholder agreements. With RAG against your standard clauses, you get deviation reports that your lawyers then review. For professionals bound by secrecy obligations: check the access route in advance under professional law, for lawyers in Germany for example with regard to section 43a of the Federal Lawyers’ Act (BRAO).

Software development with Claude Code

Claude Code is Anthropic’s coding agent for the terminal and IDE, from refactoring entire repositories to autonomous tasks. Enterprises run it with commercial access via the Claude API, AWS Bedrock or the Gemini Enterprise Agent Platform and set budgets per team via the AI Gateway. How we use AI agents in our own development is shown in our article on AI-assisted software development with OpenCode.

Research and synthesis

For synthesis tasks with multiple sources and weighted conclusions, Claude delivers traceable summaries, for example for market analyses, patent reviews or study evaluations. Always ask for source references and spot-check them.

Customer service

For high volumes in support, Haiku 5.5 is a good fit; for demanding requests in B2B service or the premium segment, Sonnet 5.5. The knowledge base should be connected via RAG so that answers are based on your own documents.

Documents, presentations and long-form content

Whitepaper drafts, study reports, slides and spreadsheets: according to Anthropic, Sonnet 5.5 is strong at documents, slides and spreadsheets. For long texts with a consistent brand voice, a style guide as a system instruction pays off.

What does Claude cost in the enterprise?

There are two billing models. Claude Team and Claude Enterprise are paid directly to Anthropic; according to Anthropic, Enterprise is also available as a usage-based plan (prices: see provider). Via the Claude API, AWS Bedrock and the Gemini Enterprise Agent Platform you pay per input and output token, tiered by model; the manufacturer’s prices are in the model table above.

Via Bedrock and the Agent Platform, current provider prices apply; tokens are billed at the hyperscaler’s list price, with no surcharge by innFactory. You’ll find the applicable token prices directly in the AWS Bedrock and Google Cloud price lists.

Per-user SaaS platforms additionally bill per seat — so their cost scales linearly with the number of users. How that plays out in the total bill is something we examine in our comparisons to Langdock, Logicc and Telekom Business GPT.

With CompanyGPT we instead use a fixed-price/maintenance model that does not scale per user — tokens are billed directly by the hyperscaler. Current prices are available on the product page or on request.

To keep token costs predictable, the AI Gateway enforces budgets and hard caps per team and member and reports Claude Code agent time per repository.

GDPR and EU AI Act: practical checklist

• Commercial access route chosen: no consumer plans (Free, Pro, Max) for business data
• DPA signed with the right party: Anthropic (Team, Enterprise, API) or the cloud provider (AWS, Google Cloud)
• EU processing explicitly configured: EU geo profile on Bedrock or EU multi-region on the Agent Platform, EU availability of the specific model version checked
• Retention clarified: covered models (30 days), need for zero data retention, retention periods on Enterprise
• Feedback feature on Team and Enterprise assessed and disabled if required
• Sub-processors and transfer impact assessment documented
• Audit logs stored audit-proof
• High-risk use cases (recruitment, credit, critical infrastructure) separately assessed (EU AI Act)
• DPIA completed for relevant applications
• Employee training under Art. 4 EU AI Act documented

If you’re not sure where you stand: our AI Compliance service delivers a full readiness report in 4 weeks.

Frequently asked questions

Which company is behind Claude? Claude is developed by Anthropic, PBC, a public benefit corporation founded in 2021 and headquartered in San Francisco; its CEO is co-founder Dario Amodei. The contracting entity for customers in the EEA, Switzerland and the United Kingdom is Anthropic Ireland, Limited in Dublin. In Germany, Anthropic is represented by Anthropic Germany GmbH and an office in Munich.

Is Claude GDPR-compliant? Claude can be used in a GDPR-compliant way, but not automatically. It requires a commercial access route (Claude Team, Claude Enterprise, the Claude API or a cloud provider) with a DPA; the consumer plans Free, Pro and Max are not suitable for business data. If you need processing in the EU, use AWS Bedrock with the EU geo profile or the EU multi-region of the Gemini Enterprise Agent Platform. Platforms such as CompanyGPT bundle this inside your own cloud tenant.

Is there a DPA with Anthropic? Yes. Claude Team, Claude Enterprise and the Claude API are covered by Anthropic’s Data Processing Addendum, which is incorporated into the Commercial Terms; the customer is the controller and Anthropic the processor. The DPA contains the EU standard contractual clauses (Modules Two and Three), an obligation to report security breaches within 48 hours and a 15-day right to object to new sub-processors. If you use Claude via AWS Bedrock or the Gemini Enterprise Agent Platform, you sign the DPA with AWS or Google Cloud instead.

Does Claude have servers in the EU? Anthropic itself offers no EU data residency: according to Anthropic, data from claude.ai, Team and Enterprise is stored in the US, and the Claude API only offers the inference locations “global” and “us”. Processing in the EU is available via AWS Bedrock (EU geo inference profile) and the Gemini Enterprise Agent Platform on Google Cloud (EU multi-region). On Bedrock, Frankfurt is not a fixed in-region location for any Claude model, and London is in the United Kingdom, i.e. outside the EU.

Is my data used for training? Not with the commercial offerings: under the Commercial Terms, Anthropic may not use customer content for training. The exception is feedback actively submitted via the thumbs button, which is stored for up to five years; owners of Team and Enterprise plans can disable this feature. On Free, Pro and Max, each user decides about training. On AWS Bedrock, according to AWS, model providers have no access to prompts and outputs, with the exception of Fable 5 and 5.1.

What is the difference between Claude Team and Claude Enterprise? Both are commercial business plans with a DPA and no training on customer content. According to Anthropic, Enterprise additionally offers custom retention periods from 30 days, the Compliance API, role-based access controls with SCIM groups, HIPAA-ready configurations and, for usage-based plans, the US-only inference option, among other things. Neither plan offers EU data residency. Anthropic publishes the prices on its own website (see provider).

Can Claude be bought on invoice in Germany? Yes, through the cloud providers. If you use Claude via AWS Bedrock or the Gemini Enterprise Agent Platform, you pay for tokens through your cloud invoice. innFactory GmbH is a cloud reseller for AWS, Microsoft Azure, Google Cloud and STACKIT and bills by invoice or direct debit, without a credit card and with all providers on one consolidated invoice. For payment options for Claude Team and Enterprise directly from Anthropic, see provider.

Can Claude be run on-premise? No. Claude is a proprietary model and only available as a cloud service: via claude.ai and the Claude API, AWS Bedrock, the Gemini Enterprise Agent Platform on Google Cloud or Microsoft Foundry. If you absolutely need on-premise, open models such as Llama, Mistral or DeepSeek are the way. More in our article on STACKIT AI Model Serving.

Which Claude version do enterprises need? Sonnet 5.5 is the default for most use cases, Opus 5.5 for deep reasoning and complex engineering, Haiku 5.5 for high-frequency simple tasks. Fable 5.1 is a covered model with mandatory 30-day retention and is only suitable for selected tasks. On multi-LLM platforms you set the default model per use case.

Is Claude Enterprise directly from Anthropic worth it? That depends on your requirements. Claude Enterprise offers a chat workplace with admin and compliance features; according to Anthropic, data is stored in the US, and there is no EU option. If you need processing in the EU, several model providers or chat data in your own cloud tenant, combine Claude via Bedrock or the Agent Platform with a platform such as CompanyGPT. A side-by-side comparison is available at CompanyGPT vs. Claude.

How do I sensibly combine Claude with GPT and Gemini? Through a multi-LLM platform that provides several models in parallel. Test the models with real tasks, then set a default model per use case and review the choice with every new model generation. Current versions and EU availability are on our model pages.

What about Claude Code for developers? Claude Code is Anthropic’s coding agent for the terminal and IDE. Because source code can be a trade secret, enterprises should only run Claude Code with commercial access via the Claude API, AWS Bedrock or the Gemini Enterprise Agent Platform and control budgets through a gateway such as the innFactory AI Gateway. Zero data retention is available for Claude Code with commercial API keys on request and on Enterprise plans. More in our article on the Claude Code source code leak.

How to start with Claude in your enterprise

1. Clarify the access route: Claude Team or Enterprise directly, Claude models via AWS Bedrock or Google Cloud, or a platform in your own tenant? We help here.
2. Set up contract and billing: DPA with Anthropic or the cloud provider, cloud account on invoice via innFactory GmbH.
3. Pilot with Sonnet 5.5: 10–30 power users, 6 weeks, clear use cases.
4. Multi-LLM and cost control: Claude, GPT and Gemini in parallel, default model per use case, budgets via the AI Gateway.
5. Compliance check: AI Compliance audit for EU AI Act readiness.
6. Rollout: structured AI training, self-service onboarding, continuous monitoring.

If you want a live demo of Claude in CompanyGPT — with your own use cases — book a 30-minute meeting here.

Sources

Primary sources, retrieved on 23 October 2026:

• Anthropic, Commercial Terms of Service: anthropic.com/legal/commercial-terms
• Anthropic, Data Processing Addendum: anthropic.com/legal/data-processing-addendum
• Anthropic, Privacy Policy: anthropic.com/legal/privacy
• Anthropic Privacy Center, server locations: privacy.claude.com/en/articles/7996890
• Anthropic Privacy Center, training and retention for commercial products: privacy.claude.com/en/articles/7996868 and privacy.claude.com/en/articles/7996866
• Claude Platform documentation, pricing and data residency: platform.claude.com/docs/en/about-claude/pricing and platform.claude.com/docs/en/manage-claude/data-residency
• Claude Platform documentation, Claude Platform on AWS and Claude Models on Google Cloud: platform.claude.com/docs/en/build-with-claude/
• Anthropic News, Series H, Long-Term Benefit Trust and European offices: anthropic.com/news
• AWS, Amazon Bedrock User Guide (Anthropic Claude model cards, data protection): docs.aws.amazon.com/bedrock/latest/userguide/
• Google Cloud, Gemini Enterprise Agent Platform, data residency: docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency
• Microsoft Learn, Claude models in Microsoft Foundry: learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/claude-models

Written by Tobias Jonas, Co-CEO innFactory AI Consulting GmbH. Date: 23 October 2026. All information is based on the providers’ public documentation as of that date; their current information prevails. This is not legal advice for individual cases. Models and availability change frequently — we update this article regularly.