Claude for Legal is not built for European or Swiss legal teams. The connector list makes the US focus clear: iManage, Everlaw, CourtListener, US e-discovery systems. Data residency in the EU or Switzerland, GDPR compliance, and attorney-client privilege requirements are not addressed in the launch announcement. Legal teams working in Zurich, Basel, or Munich need to know this now.
What Anthropic announced with Claude for Legal
According to Anthropic's launch announcement, Claude for Legal is a publicly available repository of system prompts, workflow configurations, and MCP connectors that can run as a Claude Cowork plugin or through the Managed Agents API. It is not a new, legally specialized model - under the hood it runs the same Opus model as every other Claude application.
The main surface is a chat interface. According to the announcement, twelve practice plugins cover areas such as commercial, corporate, employment, privacy, litigation, regulatory, IP, and AI governance - each with a cold-start interview that ingests the team's own playbook.
The ecosystem includes more than 20 connectors to third-party systems, among them iManage, NetDocuments, Ironclad, DocuSign, Everlaw, Relativity, Box, Thomson Reuters, and CourtListener. Named agents - preconfigured workflows such as a Vendor Agreement Reviewer and an NDA Triager - are available out of the box. Freshfields is publicly named as an early adopter, with a rollout across 33 offices.
The connector list as a map: built for the US legal market
One look at the connector list is enough to understand the target audience. CourtListener is a database of US court decisions. Everlaw is a US e-discovery platform. iManage and NetDocuments are document management systems that are primarily used in US and UK large law firms.
This is a description, not a criticism. Anthropic is building for its largest market: AmLaw 100 firms, US litigation teams, US in-house functions with established US software stacks. The launch makes very good sense as a US product.
For a legal team in Basel, Frankfurt, or Zurich, the reality looks different. The systems in use, the legal frameworks, and the regulatory requirements differ fundamentally. Connectors for systems that German mid-sized companies or Swiss businesses actually use are absent from the announcement entirely.
What is missing for the European reality
Four gaps stand out immediately.
Data residency and third-country transfers. The launch announcement does not address EU or Swiss data residency. For legal teams working with client-related data, this is not a formality. The GDPR sets clear requirements for processor arrangements and third-country transfers; Art. 16 revDSG requires appropriate safeguards for disclosures to countries without adequate protection. Where Anthropic processes data, and whether a DPA can be concluded that satisfies GDPR Art. 28 and revDSG Art. 9, remains open.
Professional secrecy and no training on client data. Swiss attorneys are subject to professional secrecy under Art. 13 BGFA; the criminal law backing comes from Art. 321 of the Swiss Criminal Code, which expressly extends to auxiliary persons. Anyone uploading client-related documents to an external system must ensure that data is not used for model training and that no unauthorized third parties can access it. Claude for Legal is silent on this in the launch announcement.
EU AI Act and compliance posture. The obligations for high-risk AI systems under Art. 6 of the EU AI Act (Annex III) were originally set to apply from August 2026; the ongoing Digital Omnibus reform is expected to push that date to December 2027 - if the reform is not adopted in time, the August 2026 date stands. How Claude for Legal fits into this framework, what transparency and documentation obligations arise, and whether Anthropic will provide corresponding compliance documentation - these questions remain unanswered.
Legal specificity for German-speaking jurisdictions. German employment law, Swiss contract law, Austrian company law - the relevant legal sources that Swiss and German-speaking legal teams use daily are not represented in the repository. CourtListener does not cover Swiss Federal Supreme Court decisions.
Why data residency and professional secrecy are not checkboxes
In practice, this means the following: an in-house legal team at a Basel pharmaceutical company that wants to analyze a vendor MSA in Claude for Legal sends contract text with supplier names, commercial terms, and potentially sensitive product information to US servers. Without a clear data processing agreement, without guaranteed data residency in the EU or Switzerland, and without a zero-data-retention commitment, this is difficult to defend legally.
That is not a theoretical point. Data protection authorities in Germany and Switzerland have repeatedly made clear that cloud services involving data transfers to the US require careful scrutiny. And for attorneys, professional secrecy applies on top of that - and cannot be overridden by the terms and conditions of a US provider.
The Swiss-U.S. Data Privacy Framework, in force since 15 September 2024, eases transfers to DPF-certified US companies. Anthropic, however, does not appear as certified on the official DPF list - so this simplified transfer route is not available for Claude for Legal.
Chat as a surface vs. end-to-end in-house processes
Anthropic's underlying assumption is that lawyers want to work in chat. That holds for many legal tasks. For in-house legal teams at large companies, it is only partly true.
In-house teams run structured processes across thousands of documents: vendor onboarding with a hundred supplier contracts per quarter, high-volume NDA processing, mass-repapering after a regulatory change, contract repositories with obligation tracking, renewal alerts and clause-level search, M&A integration where the contracts of an acquired entity need to be absorbed into the live system. These workflows require structure, audit trails, role-based access, and persistence.
A chat thread does not deliver that. Not because Claude is a poor model - but because the surface is not designed for this process architecture. Vendor Agreement Reviewer and NDA Triager exist as named agents in Claude for Legal, but as chat workflows, not as scalable processes embedded in a contract lifecycle management infrastructure.
Anyone who wants to analyze a single contract is well served by a chat interface. Anyone who needs to review two hundred supplier contracts for deviating liability clauses and send the result as an Excel export to the CFO needs a different architecture.
Where CASUS fits for Swiss and European legal teams
CASUS is a Swiss legal AI platform with hosting in Switzerland and the EU, no data transfer to the US, zero data retention, and no human review (abuse monitor opt-out). These are not marketing claims - they are the prerequisites that access for attorneys and regulated companies in Switzerland and the EU requires.
CASUS is not tied to a single model or provider: it picks the best model for each use case - Claude, GPT, or another - specialized for the way lawyers actually work.
For the Basel pharma in-house case sketched above, the work looks like this in practice: hundreds of supplier contracts are uploaded to the AI Data Room. The team defines via prompt which fields to extract - liability clauses, termination periods, IP provisions, data protection passages. The result is a structured table that flags deviations and prioritizes by risk. No chat thread, no copy-paste.
The Risk & Quality Review is party-aware: it identifies which side a contract favors and delivers structured findings with severity ratings and concrete drafting suggestions that can be applied directly in Word. For NDA volume or standard contract types such as SPA or DPA, the Benchmark shows how far a document deviates from the team's own playbook or an established market standard - with a percentage score and clause suggestions.
The legal research function searches across more than 660,000 cantonal and federal court decisions as well as statutory articles. The relevant reasoning sections are displayed inline, without leaving the workflow. That is the difference from CourtListener for a team in Zurich or Geneva.
A full description of the technical and legal foundations of data security is available on the security page.
For legal teams that want to start with the platform: CASUS offers a free entry point at app.getcasus.com/signup. Setup takes a few minutes, the Word add-in is ready to use immediately, and the team is available for structured onboarding conversations.
Comparison: Claude for Legal vs. CASUS
Criterion | Claude for Legal | CASUS |
|---|---|---|
Target market | US AmLaw, UK large firms | Swiss law firms, CH/EU in-house |
Data residency | Not addressed in launch | Switzerland / EU |
Training on client data | No commitment in launch | Zero Data Retention |
Professional secrecy / DSG | Not addressed | No human review, no US transfer |
EU AI Act compliance posture | Not discussed | Hosting EU/CH; no US transfer |
Connectors / surfaces | iManage, NetDocuments, Everlaw, Relativity, CourtListener, TR (20+) | Word add-in, web app |
In-house process support | Chat/Word-based named agents | AI Data Room (structured mass extraction) |
Swiss legal sources | Not included | 660,000+ cantonal and federal decisions |
Model | Claude Opus 4.7 (generic) | Best model per use case (Claude, GPT, etc.), specialized for legal work |
FAQ
Is Claude for Legal available in Europe and Switzerland?
According to Anthropic's launch announcement, Claude for Legal is broadly accessible, but the platform does not address EU or Swiss data residency, a GDPR-compliant processor structure, or attorney professional secrecy. Legally defensible use by European law firms or regulated companies requires a careful review of these points.
Does Claude for Legal store or train on client data?
The launch announcement contains no clear commitment on this point. For Swiss attorneys and in-house teams working under Art. 13 BGFA and the revDSG, this is an open question that needs to be resolved before any productive use.
Does Claude for Legal satisfy revDSG, GDPR, and professional secrecy obligations?
Based on the current launch announcement, not demonstrably. Processor arrangements under revDSG Art. 9 and GDPR Art. 28 require a binding contract with a provider that hosts in the EU or Switzerland, or provides equivalent safeguards. Claude for Legal makes no public statement on this.
Claude for Legal vs. Harvey vs. CASUS - what is the difference?
Harvey also targets primarily US large firms and UK offices, but offers dedicated deployments for certain European markets. Claude for Legal is a plugin ecosystem on a generic model, built around US software stacks. CASUS is a Swiss platform with hosting in Switzerland and the EU, zero data retention, no US data transfer, and Swiss legal sources - built for CH/EU legal teams that need a compliant solution today.
Is Claude for Legal suitable for in-house legal teams?
For one-off contract analysis and chat-based research, yes. For structured, scalable in-house processes across large document volumes - vendor onboarding, mass-repapering, clause-matrix extraction - a platform with bulk processing, audit trails, and persistent data management is needed. Claude for Legal is designed around chat workflows, not this process architecture.
Which systems does Claude for Legal integrate with?
According to the launch announcement, more than 20 connectors, among them iManage, NetDocuments, Ironclad, DocuSign, Everlaw, Relativity, Box, Thomson Reuters, and CourtListener. These are uniformly US- or UK-focused legal tech systems. No connectors to typical German or Swiss document management systems or legal source databases have been announced.
Is Claude for Legal a dedicated, legally specialized model?
No. According to Anthropic's announcement, Claude for Legal runs on the same Opus model as all other Claude applications. The "specialization" comes from system prompts, workflow configurations, and connectors - not from a model separately trained on legal vocabulary or jurisdiction-specific case law.
For which Swiss use cases is CASUS better suited?
Anywhere data residency, professional secrecy, and Swiss legal sources matter: contract analysis in M&A processes, high-volume NDA review, due diligence extraction across large document sets, legal research with Federal Supreme Court decisions and cantonal rulings, and proofreading before sending contracts and briefs. The AI Chat with Agent Mode covers ad-hoc analysis and direct document editing - all without US data transfer.
