Server location, processing, sub-processors: the three questions
"Servers in Germany" answers only the first of three questions to put to an AI phone assistant. The second: where do speech recognition, the language model and speech output run, at which company, and on what basis for transfers to the US? The third: who is involved as a sub-processor? A public sub-processor list answers the third fastest.
This article is orientation, not legal advice. For your specific situation the topic belongs with your data protection adviser.
Server location is not processing location
An easy mistake when assessing a provider: "servers in Germany" is read as a complete statement about the path the data takes. In fact the server location only describes where the application and the stored data sit. A phone call with an AI, though, passes through several stations: the audio goes into speech recognition, the recognised text into a language model, the reply into speech output. Each of those stations can run somewhere other than the advertised server — and that is where it is decided who processes your callers' voice and words. Which is why it pays to ask about the whole data path, not just the location.
Question 1: Where is it hosted — and by whom?
The basic question remains a fair one: which data centre holds the application, the transcripts and the contact details, and who runs it? Precision counts here, not brochure language. "German cloud" can mean many things; an answer that names the operator and the site is one you can rely on. It is worth following up on the operating model too: does the provider run its application itself on rented servers, or does it use a ready-made platform that in turn pulls in further services? The shorter the chain, the fewer parties you have to check.
Question 2: Where does the processing run — including speech recognition, the language model and speech output?
This is the question on which providers differ. Feel free to put it in technical terms: "which endpoints do speech recognition, the language model and speech output run on — EU or US? Which company operates them?" Some US providers of speech AI offer EU endpoints; a provider that uses them can name the region and the transfer basis for each station. Evasive answers ("it all runs in the cloud, it's all GDPR compliant") are a finding in themselves. Keep asking until you know the company, the region and the basis for every station.
An EU endpoint means the data is processed on servers in the EU. A US company, though, remains a US company. Under US law, providers of communication and cloud services must, when ordered by the authorities, also disclose data held outside the United States, provided it is in their possession, custody or control (18 U.S.C. § 2713, introduced in 2018 by the CLOUD Act). How far that reaches is not settled: the European Data Protection Board (EDPB) and the European Data Protection Supervisor already named open questions in 2019, for example on what "control" means for EU subsidiaries of US groups. The EDPB counts such access risks as part of checking whether a processor provides sufficient guarantees (Article 28(1) GDPR) — even where the data is processed in the EU (Guidelines 05/2021, section 2.3). How much weight that carries for your calls is something to settle case by case with your data protection adviser.
Transfers to the US: what they rest on
Personal data may only be transferred to a country outside the EU and the EEA if the conditions of Chapter V GDPR are met (Article 44). Possible bases include, in particular, an adequacy decision of the European Commission (Article 45) or appropriate safeguards such as the Commission's standard contractual clauses (Article 46(2)(c)). For the US, an adequacy decision has applied since 10 July 2023, the EU-US Data Privacy Framework; it covers only companies certified under it.
Whether that decision will hold is an open question. The General Court of the EU dismissed an action against it on 3 September 2025 (Case T-553/23, Latombe); the appeal against that judgment is pending before the Court of Justice (C-703/25 P) and, as far as we know on 15 November 2026, has not been decided. In addition, on 31 July 2026 the EDPB asked the Commission to assess whether a US Supreme Court judgment on the independence of the US Federal Trade Commission affects the decision. So for every transfer to the US, ask what it rests on and what applies if that basis falls away. Where it rests on standard contractual clauses, the exporter must also assess whether the law and practice of the destination country undermine the protection (CJEU, "Schrems II" judgment, C-311/18; EDPB Recommendations 01/2020) — ask whether that assessment is documented.
Question 3: Who is involved — the sub-processor list
Every service provider that processes your callers' personal data on the provider's behalf is a further processor, a sub-processor for short. The provider may only engage one with your prior authorisation, specific or general; under a general authorisation it must inform you of every intended change so that you can object (Article 28(2) GDPR). The EDPB goes further: controllers should have the identity of all processors and sub-processors — name, address, contact person — readily available at all times, and the processor should supply it proactively and keep it up to date (Opinion 22/2024, paras 31–32).
The form in which a provider supplies this says a lot: a public list giving the provider, the task, the registered office and the basis for each transfer shows that it expects scrutiny from data protection officers. A list "on request" is the middle way. A blanket clause without a list ("we use carefully selected service providers") does not give you the identities the EDPB expects — ask for the names before you sign. Where the privacy notice and the sub-processor list describe the same station, speech recognition for example, they should name the same service providers.
Why this matters when buying — not just legally
The three questions are also a market question: for 88 % of companies the AI provider's country of origin is important; "in this context, 93 per cent would prefer AI from Germany" (Bitkom, September 2025, 604 companies in Germany with 20 or more employees) — whether the 93 % refers to all respondents or to those 88 %, the press release leaves open. So your own business customers may ask these questions too, their data protection officers at the latest. A business that can answer them for its AI phone assistant is prepared.
scoco.ai answers them like this: servers, database and backups in Germany (Hetzner, Falkenstein); speech recognition and the language model in the EU, at providers that are US companies or belong to a US group; speech output can run through a US provider that receives only the assistant's reply texts, no caller audio. Everyone involved is listed in the public sub-processor list, each with task, registered office or processing region, the data processed and the transfer basis; it does not give addresses or contact persons. scoco.ai announces changes at least 30 days in advance by email; customers can object within 14 days on an important data protection ground (data processing agreement, § 6). As the basis for transfers to third countries, the data processing agreement names an adequacy decision such as the Data Privacy Framework or standard contractual clauses; for DPF-certified providers it keeps standard contractual clauses in reserve as a fallback, and customers can inspect the transfer impact assessment for transfers based on standard contractual clauses on request (data processing agreement, § 12).
Honestly: the location is not everything
Two qualifications belong here. First: a German server location replaces neither a deletion policy nor documented technical and organisational measures (Article 32 GDPR); whether calls are recorded as audio is a separate question to check. Second: the GDPR does not prohibit US companies in the chain — Chapter V sets out the conditions under which transfers are permitted. What matters is that every transfer has a basis under Chapter V and that the arrangement is disclosed. So test absolute statements such as "your data never leaves Germany" against the sub-processor list: they have to hold for every station, including speech recognition, the language model and speech output.
Frequently asked questions
What exactly is a sub-processor?
A service provider that your processor in turn engages to process personal data on your behalf — for speech recognition, speech output or hosting, for example. What counts for you is the whole chain: the provider may only engage sub-processors with your prior authorisation and, under a general authorisation, must inform you of every intended change so that you can object (Article 28(2) GDPR).
Is "servers in the EU" enough instead of "servers in Germany"?
For the GDPR it makes no fundamental difference: it applies throughout the EU and the EEA, and its rules on transfers (Articles 44 et seq. GDPR) concern only countries outside them. Professional rules can still matter: German lawyers, for example, may in principle give a service provider that performs its service abroad — including elsewhere in the EU — access to client secrets only if the protection of secrets there is comparable to that in Germany (§ 43e(4) BRAO, the German Federal Lawyers' Act). Whether something similar applies to your profession is a question for your professional chamber. Whatever the country, hosting, processing and everyone involved should be named in full.
Are US sub-processors with an EU endpoint unproblematic?
There is no blanket answer. The GDPR permits transfers to the US, for example under the adequacy decision on the EU-US Data Privacy Framework for certified companies or with standard contractual clauses (Article 45, Article 46(2)(c) GDPR); whether that decision will stand before the Court of Justice is open (as of 15 November 2026). An EU endpoint keeps the data on servers in the EU, but does not change the fact that US law such as the CLOUD Act can apply to the company. How you weigh that is a case-by-case assessment for your data protection adviser; your job when choosing a provider is to know the arrangement in full.
How do I keep the sub-processor list up to date?
The duty to inform lies with the provider: under a general authorisation it must inform you of every intended change (Article 28(2), second sentence, GDPR), and in the view of the European Data Protection Board it should keep the information up to date of its own accord (Opinion 22/2024). Check in the data processing agreement how, and with what notice, changes are announced and how you object, and archive the state of the list when you sign.
Status & sources
As of 2026-11-15.
- Not legal advice. Sources: GDPR Article 28(1) and (2), Article 32, Article 44, Article 45(1), Article 46(2)(c)
- Commission Implementing Decision (EU) 2023/1795 of 10 July 2023 on the adequate level of protection of personal data under the EU-US Data Privacy Framework
- General Court, judgment of 3 September 2025, T-553/23, Latombe v Commission: action for annulment of Implementing Decision (EU) 2023/1795 dismissed (press release No 106/25)
- Appeal C-703/25 P, brought on 31 October 2025 by Philippe Latombe against judgment T-553/23 (OJ C/2025/6610); not decided as far as we know on 15 November 2026
- European Data Protection Board, letter from the Chair to Commissioner McGrath of 31 July 2026 on the US Supreme Court judgment in Trump v. Slaughter of 29 June 2026 (Ref. Ares(2026)7540711): asks the Commission to assess the effect on Implementing Decision (EU) 2023/1795
- CJEU, judgment of 16 July 2020, C-311/18, Data Protection Commissioner v Facebook Ireland and Schrems ("Schrems II")
- European Data Protection Board, Recommendations 01/2020 on measures that supplement transfer tools, version of 18 June 2021
- European Data Protection Board, Guidelines 05/2021 on the interplay between the application of Article 3 and Chapter V GDPR, version 2.0 of 14 February 2023, section 2.3 (a processor subject to third-country law: access requests from authorities there count when assessing sufficient guarantees under Article 28(1))
- 18 U.S.C. § 2713 (Required preservation and disclosure of communications and records), introduced by the CLOUD Act of 23 March 2018
- European Data Protection Board and European Data Protection Supervisor, joint response to the LIBE Committee on the impact of the US CLOUD Act, 12 July 2019, annex "Initial legal assessment" (open questions on its reach, including what "control" means for EU affiliates of US companies)
- European Data Protection Board, Opinion 22/2024 on certain obligations following from the reliance on processor(s) and sub-processor(s), adopted 7 October 2024, paras 31–32
- § 43e(4) Bundesrechtsanwaltsordnung (BRAO, German Federal Lawyers' Act): services provided abroad
- Bitkom, press release of 15 September 2025 (604 companies in Germany with 20 or more employees, by telephone, weeks 27 to 32 of 2025): for 88 % of companies the AI provider's country of origin is important; "in this context, 93 per cent would prefer AI from Germany" (the base of the 93 % is not stated)
- scoco.ai product information: as of 15 November 2026 (sub-processor list version 0.7 of 30 October 2026; data processing agreement §§ 6 and 12, as of 15 November 2026)