Selling data to AI companies from Germany
Of the 24 buyers of company records whose pages we read, 5 say they take holders from anywhere, work with holders abroad or list Germany, 2 take holders abroad but rank the US first, 2 would not take a holder here, and 15 say nothing about where a holder can be based. Buyers' own estimators value a company here at 0.71 times the same company in the US. The US withholds 0% of a licence fee paid to a holder here under its tax treaty, when the holder claims it. Before a licence, the GDPR asks some work work of the holder: a written legitimate interest assessment and compatibility test, notice with a chance to object, and a buyer on the Privacy Framework list or contract clauses; de-identifying first removes most of it. This is research from public sources, not legal or tax advice.
- Privacy work before a licence
- Some work (the GDPR)
- US withholding on a licence fee
- 0% under the treaty
- Price against a US company
- 0.71 times
- We sell for holders here
- Not yet
What the GDPR asks before a licence
Records about a company alone are outside privacy law. Every record that names or can identify a person (a customer, a member of staff, a caller, someone named in an email) is inside it, and these are the points that decide whether it can be licensed. Our reading first, then the law in its own words.
Using records for a new purpose
A licence to an AI buyer is a new purpose. In practice the holder relies on legitimate interests, writes down the three-part test for each lot and a compatibility test, tells people before it starts and lets them object, and leaves out health, union, belief and similar data. A company holding its customers' records as their processor cannot license them at all.
The 9 sources it rests on, in their own words
- GDPR Article 5(1)(b): purpose limitation (Regulation (EU) 2016/679, Art 5(1)(b))
“collected for specified, explicit and legitimate purposes and not further processed in a manner that is incompatible with those purposes; further processing for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes shall, in accordance with Article 89(1), not be considered to be incompatible with the initial purposes”
- GDPR Article 6(4): compatibility test for a new purpose (Regulation (EU) 2016/679, Art 6(4))
“the controller shall, in order to ascertain whether processing for another purpose is compatible with the purpose for which the personal data are initially collected, take into account, inter alia: (a) any link between the purposes for which the personal data have been collected and the purposes of the intended further processing; (b) the context in which the personal data have been collected, in particular regarding the relationship between data subjects and the controller; (c) the nature of the personal data, in particular whether special categories of personal data are processed, pursuant to Article 9”
- GDPR Article 6(1)(f): legitimate interests as a lawful basis (Regulation (EU) 2016/679, Art 6(1)(f))
“processing is necessary for the purposes of the legitimate interests pursued by the controller or by a third party, except where such interests are overridden by the interests or fundamental rights and freedoms of the data subject which require protection of personal data, in particular where the data subject is a child.”
- Koninklijke Nederlandse Lawn Tennisbond (C-621/22), para 51: ask the people first as the less intrusive means (Case C-621/22, Koninklijke Nederlandse Lawn Tennisbond v Autoriteit Persoonsgegevens, EU:C:2024:857 (4 October 2024), para 51)
“it would, in particular, be possible for a sports federation such as the KNLTB, wishing to disclose its members' personal data to third parties for consideration, to inform its members beforehand and to ask them whether they want their data to be transmitted to those third parties for advertising or marketing purposes.”
- EDPB Opinion 28/2024, para 66: the legitimate interest test must be assessed and documented (EDPB Opinion 28/2024, adopted 17 December 2024, para 66)
“SAs should verify that controllers have carefully assessed and documented whether the three following cumulative conditions are met: (i) the pursuit of a legitimate interest by the controller or by a third party; (ii) the processing is necessary to pursue the legitimate interest; and (iii) the legitimate interest is not overridden by the interests or fundamental rights and freedoms of the data subjects”
- GDPR Article 21(1): right to object (Regulation (EU) 2016/679, Art 21(1))
“The data subject shall have the right to object, on grounds relating to his or her particular situation, at any time to processing of personal data concerning him or her which is based on point (e) or (f) of Article 6(1), including profiling based on those provisions. The controller shall no longer process the personal data unless the controller demonstrates compelling legitimate grounds for the processing which override the interests, rights and freedoms of the data subject or for the establishment, exercise or defence of legal claims.”
- GDPR Article 13(3): telling people before a new purpose (Regulation (EU) 2016/679, Art 13(3))
“Where the controller intends to further process the personal data for a purpose other than that for which the personal data were collected, the controller shall provide the data subject prior to that further processing with information on that other purpose and with any relevant further information as referred to in paragraph 2.”
- GDPR Article 9(1): special categories of personal data (Regulation (EU) 2016/679, Art 9(1))
“Processing of personal data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, or trade union membership, and the processing of genetic data, biometric data for the purpose of uniquely identifying a natural person, data concerning health or data concerning a natural person's sex life or sexual orientation shall be prohibited.”
- GDPR Article 28(10): a processor that decides purposes becomes a controller (Regulation (EU) 2016/679, Art 28(10))
“if a processor infringes this Regulation by determining the purposes and means of processing, the processor shall be considered to be a controller in respect of that processing.”
Sending personal data to a buyer in the US
Identifiable personal data can go to a US buyer on the EU-US Data Privacy Framework only if the buyer is on its list; otherwise the two sign the standard contractual clauses, and the licence names AI model training as the purpose. Truly anonymous data is not a transfer at all.
The 6 sources it rests on, in their own words
- GDPR Article 44 (general principle for transfers) (Regulation (EU) 2016/679, Art 44)
“Any transfer of personal data which are undergoing processing or are intended for processing after transfer to a third country or to an international organisation shall take place only if, subject to the other provisions of this Regulation, the conditions laid down in this Chapter are complied with by the controller and processor”
- GDPR Article 46(1): appropriate safeguards for a transfer without adequacy (Regulation (EU) 2016/679, Art 46(1))
“In the absence of a decision pursuant to Article 45(3), a controller or processor may transfer personal data to a third country or an international organisation only if the controller or processor has provided appropriate safeguards, and on condition that enforceable data subject rights and effective legal remedies for data subjects are available.”
- EU-US Data Privacy Framework adequacy decision, Article 1 (Commission Implementing Decision (EU) 2023/1795, Art 1)
“For the purpose of Article 45 of Regulation (EU) 2016/679, the United States ensures an adequate level of protection for personal data transferred from the Union to organisations in the United States that are included in the 'Data Privacy Framework List', maintained and made publicly available by the U.S. Department of Commerce, in accordance with Section I.3 of Annex I.”
- Latombe v Commission, judgment of the General Court (Case T-553/23, Latombe v Commission, EU:T:2025:831 (3 September 2025), para 204 and operative part)
“In the light of all those elements, the fifth plea in law must be rejected and, accordingly, the action must be dismissed in its entirety. ... Dismisses the action;”
- Standard contractual clauses, Module One (controller to controller), Clause 8.1: purpose limitation (Commission Implementing Decision (EU) 2021/914, Annex, Module One, Clause 8.1)
“The data importer shall process the personal data only for the specific purpose(s) of the transfer, as set out in Annex I.B. It may only process the personal data for another purpose: (i) where it has obtained the data subject's prior consent;”
- EDPB Guidelines 05/2021: when a disclosure is a 'transfer' under Chapter V (EDPB Guidelines 05/2021 on the interplay between Article 3 and Chapter V, Version 2.0, adopted 14 February 2023, executive summary)
“1) A controller or a processor ("exporter") is subject to the GDPR for the given processing. 2) The exporter discloses by transmission or otherwise makes personal data, subject to this processing, available to another controller, joint controller or processor ("importer"). 3) The importer is in a third country, irrespective of whether or not this importer is subject to the GDPR for the given processing in accordance with Article 3, or is an international organisation.”
Anonymised and de-identified data
Truly anonymous information is outside the GDPR, but removing names is not enough: no one may be singled out, linked or inferred. The holder, which keeps the originals, still treats the lot as personal data; a buyer that receives a de-identified lot may not.
The 5 sources it rests on, in their own words
- GDPR Recital 26 (anonymous information) (Regulation (EU) 2016/679, recital 26)
“The principles of data protection should therefore not apply to anonymous information, namely information which does not relate to an identified or identifiable natural person or to personal data rendered anonymous in such a manner that the data subject is not or no longer identifiable. This Regulation does not therefore concern the processing of such anonymous information”
- Article 29 Working Party Opinion 05/2014: removing direct identifiers is not enough (Article 29 Working Party, Opinion 05/2014 on Anonymisation Techniques (WP216), adopted 10 April 2014, section 2.2.2)
“An effective anonymisation solution prevents all parties from singling out an individual in a dataset, from linking two records within a dataset (or between two separate datasets) and from inferring any information in such dataset. Generally speaking, therefore, removing directly identifying elements in itself is not enough to ensure that identification of the data subject is no longer possible.”
- EDPS v SRB, para 86 (Case C-413/23 P, EDPS v SRB (4 September 2025), para 86)
“pseudonymised data must not be regarded as constituting, in all cases and for every person, personal data ... in so far as pseudonymisation may, depending on the circumstances of the case, effectively prevent persons other than the controller from identifying the data subject in such a way that, for them, the data subject is not or is no longer identifiable.”
- EDPS v SRB, para 111 (Case C-413/23 P, EDPS v SRB (4 September 2025), para 111)
“for the purposes of applying the obligation to provide information laid down in Article 15(1)(d) of Regulation 2018/1725, the identifiable nature of the data subject must be assessed at the time of collection of the data and from the point of view of the controller.”
- EDPB Guidelines 01/2025 on Pseudonymisation (version for public consultation) (EDPB Guidelines 01/2025, adopted 16 January 2025, executive summary)
“Pseudonymised data, which could be attributed to a natural person by the use of additional information, is to be considered information on an identifiable natural person ... This statement also holds true if pseudonymised data and additional information are not in the hands of the same person.”
Rules on AI training and data mining
The European Data Protection Board accepts legitimate interests for AI training when the test is written down, with masking and an opt-out as safeguards, and expects a buyer to check where its data came from. The AI Act makes a model's provider publish a general summary of its training data. The text and data mining exception does not reach private records, which is why they need a licence.
The 9 sources it rests on, in their own words
- EDPB Opinion 28/2024, para 66: the legitimate interest test must be assessed and documented (EDPB Opinion 28/2024, adopted 17 December 2024, para 66)
“SAs should verify that controllers have carefully assessed and documented whether the three following cumulative conditions are met: (i) the pursuit of a legitimate interest by the controller or by a third party; (ii) the processing is necessary to pursue the legitimate interest; and (iii) the legitimate interest is not overridden by the interests or fundamental rights and freedoms of the data subjects”
- EDPB Opinion 28/2024, para 101: masking and substituting personal data (EDPB Opinion 28/2024, adopted 17 December 2024, para 101(c))
“Measures to mask personal data or to substitute it with fake personal data in the training set ... This measure may be particularly appropriate when the actual substantive content of the data is not relevant to the overall processing (e.g. in LLM training).”
- EDPB Opinion 28/2024, para 129: the next party must check where the data came from (EDPB Opinion 28/2024, adopted 17 December 2024, para 129)
“Such evaluation by SAs should take into account whether the controller has assessed some non-exhaustive criteria, such as the source of the data and whether the AI model is the result of an infringement of the GDPR, particularly if it was determined by a SA or a court, so that the controller deploying the model could not ignore that the initial processing was unlawful.”
- AI Act, Article 53(1)(d): public summary of training content (Regulation (EU) 2024/1689, Art 53(1)(d))
“draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office.”
- Commission template for the public summary of AI training content: how licensed and private datasets are described (European Commission AI Office, Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models, 24 July 2025, recital (19) and Section 2.2.2)
“limited disclosure of information is required for licensed data given that the rightsholders concerned are parties to the licensing agreements (see Section 2.2.1 of the Template). Furthermore, private datasets not commercially licensed by rightsholders and obtained from other third parties have to be listed only if publicly known (or the provider wants to make them publicly known), and otherwise described in a general manner (see Section 2.2.2 of the Template).”
- DSM Directive 2019/790, Article 4: the text and data mining exception and its opt-out (Directive (EU) 2019/790, Art 4(1) and (3))
“Member States shall provide for an exception or limitation to the rights provided for in Article 5(a) and Article 7(1) of Directive 96/9/EC, Article 2 of Directive 2001/29/EC, Article 4(1)(a) and (b) of Directive 2009/24/EC and Article 15(1) of this Directive for reproductions and extractions of lawfully accessible works and other subject matter for the purposes of text and data mining. ... The exception or limitation provided for in paragraph 1 shall apply on condition that the use of works and other subject matter referred to in that paragraph has not been expressly reserved by their rightholders in an appropriate manner, such as machine-readable means in the case of content made publicly available online.”
- Irish DPC statement on Meta AI (21 May 2025): the safeguards it asked for (Data Protection Commission (Ireland), 'DPC statement on Meta AI', 21 May 2025)
“Updated its measures to protect data subjects, such as de-identification, filtering of data sets and output filters; and Updated its risks assessments and other required documentation under GDPR such as their Legitimate Interest Assessment, Data Protection Impact Assessment and a Compatibility Assessment.”
- Irish DPC guidance on AI, large language models and data protection (18 July 2024) (Data Protection Commission (Ireland), 'AI, Large Language Models and Data Protection', 18 July 2024)
“If you are creating an AI model with existing personal data you have collected, then you also need to consider if any proposed processing purposes are within the scope of the existing legal basis you have. ... If you have data sharing agreements with other organisations allowing their personal data to be used for training your model and sharing or licensing that model, you must ensure that you have a legal basis to process that personal data in that way, as well as ensuring that the processing is fair and transparent.”
- OLG Koln, 15 UKl 2/25 (23 May 2025): economic interests can be legitimate interests for AI training (Oberlandesgericht Koln, Urteil vom 23.05.2025, 15 UKl 2/25 (ECLI:DE:OLGK:2025:0523.15UKL2.25.00), para 59)
“Als berechtigte Interessen kommen neben rechtlichen und ideellen insbesondere auch wirtschaftliche Belange in Betracht (EuGH, Urteil vom 4. Oktober 2024 - C-621/22 -, juris, Rn. 47; Hüger, ZfdR 2024, 263, 272).”
Tax on the licence fee
What the US withholds
A licence fee for records used in the US is a US-source royalty. Under the US treaty with Germany, the US withholds nothing on know-how and copyright royalties paid to a resident who claims the treaty on Form W-8BEN-E before payment; without the form the buyer must withhold 30 percent. The buyer still files a Form 1042-S.
The 4 sources it rests on, in their own words
- IRS Tax Treaty Table 1, royalty row for Germany (IRS, Table 1 (Rev. May 2023), Germany row, royalty columns (income codes 10 to 12))
“GM 0 18(5) / PVIII n/a u 0 0 0 0 12(1)”
- US-Germany income tax convention, Article 12: royalties taxable only in the country of the recipient (Convention between the United States and Germany (1989) as amended, Art. 12(1)-(2))
“Royalties derived and beneficially owned by a resident of a Contracting State shall be taxable only in that State. ... payments of any kind received as a consideration for the use of, or the right to use, any copyright of a literary, artistic, or scientific work (but not including cinematographic films, or works on film, tape, or other means of reproduction for use in radio or television broadcasting); for the use of, or the right to use, any patent, trademark, design or model, plan, secret formula or process, or other like right or property; or for information concerning industrial, commercial, or scientific experience.”
- IRC section 1442(a): the same withholding duty applies to payments to foreign corporations (26 U.S.C. 1442(a))
“In the case of foreign corporations subject to taxation under this subtitle, there shall be deducted and withheld at the source in the same manner and on the same items of income as is provided in section 1441 a tax equal to 30 percent thereof.”
- IRS Instructions for Form W-8BEN-E: the tax number and the treaty claim in Part III, lines 14a and 14b (IRS, Instructions for Form W-8BEN-E (Rev. 10-2021), Line 9b and Part III, Lines 14a-14b)
“You must also provide a U.S. TIN if you are: ... Claiming benefits under an income tax treaty and have not provided a FTIN on line 9b. ... If you are claiming a reduced rate of, or exemption from, withholding under an income tax treaty you must enter the country where you are a resident for income tax treaty purposes and check the box to certify that you are a resident of that country. Line 14b. If you are claiming a reduced rate of, or exemption from, withholding under an income tax treaty you must check the box to certify that you: Derive the item of income for which the treaty benefit is claimed, and Meet the limitation on benefits provision contained in the treaty, if any.”
Tax in Germany on licensing to a US buyer
A licence to a business established in the US is supplied where the buyer is established, so the holder's invoice carries no EU VAT; keep proof that the buyer is a business.
The law it rests on, in their own words
- VAT Directive, Article 44 (general B2B place-of-supply rule) (Council Directive 2006/112/EC, Art 44 (consolidated 1 January 2025))
“The place of supply of services to a taxable person acting as such shall be the place where that person has established his business.”
Which buyers take holders from Germany
Each buyer's own rules, read on its own pages. “Says nothing about where” is not a yes: ask the buyer in writing before you count on it.
Buyers of company records
Ooak Data
Works with holders abroad. It runs a French entity for EU partners with EU-only processing and says it checks jurisdiction during scoping, but it publishes no country list, currency or tax rule.
Its 3 rules, in its own words
- Where a holder can be based
“We check jurisdiction and requirements during scoping.”
- Where a holder can be based
“Data belonging to partners located in the European Union is processed and stored exclusively within the European Union. Our European operations are conducted through Ooak Data SAS , registered with the Paris Trade and Companies Register under number 932 867 963 (RCS Paris), with its registered office at 60, rue François Ier, 75008 Paris, France.”
- How a holder abroad is paid
“Deals to date range from $20K, the smallest closed, to several hundred thousand dollars.”
- Where a holder can be based
SuperAnnotate
Takes holders from anywhere. Its data partnerships FAQ says any size, any industry, anywhere, though it publishes nothing on language, tax forms, sanctions or payout currency.
Its 2 rules, in its own words
- Where a holder can be based
“Any size, any industry, anywhere. You need records built up over time and the rights to license them.”
- How a holder abroad is paid
“We prepare the data and license it to buyers you approve. Payment on delivery, and on each relicense.”
- Where a holder can be based
The General Data Company (General Data Labs, Inc.)
Works with holders abroad. It is a San Francisco company that courts holders in Germany (conference, German, French and Spanish pages, a euro calculator) and sets no US-only rule, though it says nothing on payout method, tax forms or sanctions.
Its 4 rules, in its own words
- Where a holder can be based
“Connect with leaders from DAX companies, leading industrial businesses and Germany’s Mittelstand.”
- Where a holder can be based
“Based in San Francisco, USA. The operating company is General Data Labs, Inc.”
- Language of the records
“Both pages are available in English and, under the /de/, /fr/ and /es/ prefixes, in German, French and Spanish.”
- How a holder abroad is paid
“350.000 € bis 2,15 Mio. €”
- Where a holder can be based
Troveo
Works with holders abroad. It lists licensors in Vietnam, the Netherlands, India, South Africa, Canada, Australia and New Zealand and tells international licensors to use Form W-8BEN, but only for its content-licensing side; its company-data page states no country rule.
Its 7 rules, in its own words
- Where a holder can be based
“Featured Data Owners ... POPS Worldwide Ho Chi Minh City, Vietnam ... ODMedia Utrecht, Netherlands ... Asian News International New Delhi, India ... Arena Holdings Johannesburg, South Africa”
- Where a holder can be based
“Arcadia Entertainment Halifax, Canada ... Nine Network Sydney, Australia ... NHNZ Worldwide Auckland, New Zealand”
- How a holder abroad is paid
“Payouts are processed through our third-party vendor, Tipalti. To receive payments from Troveo, you'll need to complete your payment setup directly in your Troveo portal.”
- How a holder abroad is paid
“There is no fixed payout schedule. Payments occur only after your content has been licensed, the client has been invoiced, and payment has been received.”
- Tax forms
“For international licensors: What Is Form W‑8BEN? How to Use It for Tax Compliance Who should use W‑8BEN (non‑US individuals) and how it differs from W‑8BEN‑E and W‑9.”
- Tax forms
“I'm not a U.S. citizen or resident. Do I need a W-9? No. Non-U.S. persons may need to complete a Form W-8BEN (or another W-8 series form). Check with Ramp for their requirements for international payees.”
- Tax forms
“Avoid backup withholding: Submitting a valid W-9 prevents Ramp from withholding a portion of your payments for taxes, which can happen if a W-9 is missing or incorrect.”
- Where a holder can be based
Turing (Project Lazarus)
Takes holders from anywhere. Its FAQ says companies can be based in most regions (strongest demand in the US, UK, Canada and the EU) and that it flags jurisdictions with data-export restrictions early.
Its 6 rules, in its own words
- Where a holder can be based
“Most regions, with strongest demand in the US, UK, Canada, and the EU. Some jurisdictions carry data-export restrictions we'll flag early.”
- Language of the records
“English-language comments & issues”
- Language of the records
“PR title and body must be in English”
- How a holder abroad is paid
“One payment. No retainer, no recurring commitment.”
- How a holder abroad is paid
“We make our final buying decision, agree on pricing, and issue payment. Repos typically pay $10k-$100k+ depending on quality and complexity.”
- Sanctions
“To violate any applicable federal, state, local, or international law or regulation, including export control laws and regulations administered by the U.S. Bureau of Industry and Security (BIS) or the Office of Foreign Assets Control (OFAC);”
- Where a holder can be based
Handshake AI Data Partnerships
Takes holders abroad, ranks the US first. No page text excludes any country and its FAQ names none, but the payout calculator's own settings price Canada and Europe at 0.75 and Other at 0.30 against 1.00 for the USA, and it asks for 20 or more employees and 3 or more years.
Its 5 rules, in its own words
- Where a holder can be based
“Country USA Canada Europe Other”
- Where a holder can be based
“Company name Company website (optional) Country Continue”
- Where a holder can be based
“defaultValue:`USA`,description:`Prices at 1.00.`},canadaLabel:{type:H.String,title:`Region 2`,defaultValue:`Canada`,description:`Prices at 0.75.`},europeLabel:{type:H.String,title:`Region 3`,defaultValue:`Europe`,description:`Prices at 0.75.`},otherLabel:{type:H.String,title:`Region 4`,defaultValue:`Other`,description:`Prices at 0.30.`}”
- Where a holder can be based
“Shown before $100,000 instead of a range when the lower estimate falls below it - outside the US only.”
- How a holder abroad is paid
“Payment terms are negotiated as part of your agreement and tied to defined delivery milestones. You’ll receive written details on the payment amounts, milestones, and schedule before any data is transferred.”
- Where a holder can be based
micro1
Takes holders abroad, ranks the US first. Global in principle, but it prioritises US companies, then the UK, Canada and other Western markets, and wants English-language data.
Its 6 rules, in its own words
- Where a holder can be based
“Yes. While the program accepts referrals globally, current demand is strongest for companies based in the United States, United Kingdom, and Canada. Companies from other regions may still be considered, but selection will depend on the specific needs of active opportunities and customer demand at the time.”
- Where a holder can be based
“We currently prioritize US companies, followed by other Western markets.”
- Where a holder can be based
“Where is your company headquartered? United States Canada Europe Other”
- Language of the records
“At this time, we are primarily looking for companies that operate in English and maintain their documentation, workflows, and knowledge bases in English. Company representatives must also be able to communicate with the micro1 team in English throughout the partnership and onboarding process.”
- Tax forms
“Each party bears its own taxes as required by applicable law, and amounts stated in an Order Form are exclusive of any applicable sales, use, or value-added tax, which Licensor will state separately on its invoice where required.”
- Sanctions
“referrers and referred candidates must not reside in countries or regions where micro1 is unable to operate due to legal, regulatory, or security requirements.”
- Where a holder can be based
Appen
Says nothing about where. Its enterprise-data page sets no country, language or tax rule for company holders and it publishes no enterprise terms (Appen is an Australian-listed company with a US headquarters, but that says nothing about who it takes).
Its own words
- How a holder abroad is paid
“Terms are agreed in writing before extraction begins, and payment is on acceptance rather than submission. Appen covers extraction, anonymisation and transfer, so there are no costs to you.”
- How a holder abroad is paid
Centific
Says nothing about where. Its enterprise-data page and contact form set no country, language, tax or sanctions rule for company holders; its privacy policy calls it a global company and its website terms use Delaware law.
Its own words
- How a holder abroad is paid
“We agree to the terms in writing before extraction begins. Payment is made when the data is accepted.”
- How a holder abroad is paid
Collinear AI
Says nothing about where. The data-acquisition page says nothing about where a holder must be based, the language of records, currency or tax forms.
Its 2 rules, in its own words
- Sanctions
“You agree to comply with applicable laws in connection with its performance hereunder, including without limitation, applicable U.S. and foreign export controls, economic sanctions, and other trade controls.”
- How a holder abroad is paid
“We confirm delivery is complete, then pay according to the contract.”
- Sanctions
Corpus Bridge
Says nothing about where. No country rule is published and its assessment form offers every country, but the terms and privacy policy could not be read.
Its 3 rules, in its own words
- Where a holder can be based
“Select a country”
- Language of the records
“Languages other than English , which are underrepresented in public training data.”
- How a holder abroad is paid
“The buyer pays on the milestones in the agreement, for example half on signature and half on delivery.”
- Where a holder can be based
FileYield
Says nothing about where. No country rule for holders, only a sanctions-law eligibility clause, Arizona law and a privacy warning to EEA, UK and Swiss users that US data protection may be lower.
Its 5 rules, in its own words
- Sanctions
“Not be barred from using the Service under U.S. or other applicable law (sanctions, embargoes, OFAC lists)”
- Language of the records
“Multi-language call data is extremely valuable for conversational AI training.”
- How a holder abroad is paid
“The transfer of data, the legal contract for the sale, and the payment occur directly between the seller and buyer”
- How a holder abroad is paid
“When payments become available, payments will be processed through Stripe.”
- Tax forms
“Taxes are your responsibility. You are responsible for any sales, use, value-added, or other taxes that may apply to your purchases or sales on the Service.”
- Sanctions
Grepped
Says nothing about where. Its holder pages set no country rule, and its referral terms pay referrers in any country except comprehensively sanctioned ones, but nothing says whether a holder outside the US is taken.
Its 3 rules, in its own words
- How a holder abroad is paid
“Payments are made in US dollars by bank transfer or Wise to an account in your name.”
- Tax forms
“You are responsible for your own taxes and will provide any tax form we reasonably require, such as a W-9 or W-8BEN, before the first payment.”
- Sanctions
“Referrers in any country may be paid, except in countries subject to comprehensive US, UK or EU sanctions, or persons named on a sanctions list.”
- How a holder abroad is paid
Halluminate
Says nothing about where. The page asks only for contact details and publishes no country rule, payment method, tax form, terms or privacy policy.
HUMXN
Says nothing about where. The company-data page sets no country rule, and its legal pages are unfilled templates (no home country or governing law), so it is unknown whether a holder outside the US is taken.
Its 4 rules, in its own words
- Where a holder can be based
“legally able to enter into contracts and to perform paid work where you are located”
- How a holder abroad is paid
“Your data goes to approved AI buyers under strict confidentiality. Your company is never named, and you're paid per licence.”
- How a holder abroad is paid
“Payment for accepted work is made through Stripe to your verified payout account, after the work has passed review, on the schedule set out in the brief [e.g. within 14 days of acceptance - confirm]. You must complete Stripe's identity and payout verification before you can be paid.”
- Tax forms
“You are responsible for your own taxes, insurance, equipment and compliance with laws that apply to your business.”
- Where a holder can be based
Labelbox (Prism)
Says nothing about where. Prism says nothing about where a holder must be based (its industry notes cite only US laws such as HIPAA, GLBA and SOX), but it promises regional data residency and its scoping form asks no country.
Its 3 rules, in its own words
- How a holder abroad is paid
“How It Works From first call to first wire in under 21 days.”
- Tax forms
“You will be responsible for paying, withholding, filing, and reporting all taxes, duties, and other governmental assessments associated with your activity in connection with the Services”
- Sanctions
“Violates any law or regulation, including, without limitation, any applicable export control laws”
- How a holder abroad is paid
Magma Intelligence Inc.
Says nothing about where. Magma publishes no rule on where a holder must be based (it is a Houston, Texas company and its form asks no country); everything else is left to a data agreement it does not publish.
Its own words
- How a holder abroad is paid
“A revenue share, paid as your data licenses. Terms are in the data agreement we sign during onboarding.”
- How a holder abroad is paid
Mercor Data
Says nothing about where. The data page names no country or language rule and says holders are paid by wire transfer; the only geographic rule found is a US sanctions representation in its general terms.
Its 3 rules, in its own words
- How a holder abroad is paid
“Payment is issued via wire transfer upon delivery of the anonymized dataset to the buyer. This typically occurs within 2-4 weeks of the signed agreement, depending on data extraction and processing time.”
- How a holder abroad is paid
“No. Mercor covers all costs associated with data extraction, anonymization, and transfer. You receive compensation with no upfront costs, setup fees, or deductions from the payment.”
- Sanctions
“is (a) listed on the List of Specially Designated Nationals and Blocked Persons or on any other list of sanctioned, prohibited, or restricted parties administered by OFAC or by any other governmental entity, or (b) located in, a national or resident of, or a segment of the government of, any country or territory for which the United States maintains trade or economic sanctions or embargoes”
- How a holder abroad is paid
Polyshares
Says nothing about where. Its public pages set no country rule for companies (only 'any company with real operating history, in any sector'), and it publishes no terms or privacy policy.
Its own words
- How a holder abroad is paid
“Payment Net 30 to 60 , once the data is approved, shared, and anonymized.”
- How a holder abroad is paid
Scale AI Data Partnerships
Says nothing about where. The page names no country, language, payment or tax rule; it asks only for established operators, preferably with 40 or more employees.
SimpleClosure (AssetHub)
Says nothing about where. AssetHub's own pages set no country rule for holders (terms and privacy policy cite GDPR), but its referral programme is limited to US residents and US-organised companies.
Its 3 rules, in its own words
- Where a holder can be based
“Each Referrer must be at least 18 years old and a resident of the United States of America.”
- Where a holder can be based
“(c) send Referral Codes to any company that is not a company organized in the United States;”
- Where a holder can be based
“Official dissolution paperwork filed in the company's state of incorporation.”
- Where a holder can be based
Sunset
Says nothing about where. Sunsethq.com could not be read (Cloudflare challenge blocks all non-browser fetches), so nothing is known about where it takes holders.
Mode (BizBuyData)
Takes holders from a listed set of places. Its referral pages take companies and nonprofits based in the United States, Canada, the United Kingdom or Australia only, and its company-data page calls US-based teams the strongest fit.
Its 7 rules, in its own words
- Where a holder can be based
“Companies and nonprofits must be based in the United States, Canada, United Kingdom, or Australia.”
- Where a holder can be based
“Companies in the United States, Canada, United Kingdom, and Australia only.”
- Where a holder can be based
“Companies with several years of operating records they own and teams primarily based in the U.S. are the strongest fit.”
- Where a holder can be based
“Where your team is based USA Canada UK Australia EU Other”
- Where a holder can be based
“US direct referrals: greater of $50,000 or 6% for your first five Qualified Deals, with higher lifetime tiers. Outside the US: 6%, no minimum.”
- How a holder abroad is paid
“Mode buys the agreed copy directly and charges no broker fee. Your written agreement sets the amount and payment schedule before export.”
- Tax forms
“The percentage is based on license fees actually paid to the company, excluding taxes and reimbursements.”
- Where a holder can be based
Nyne
US holders only. Its current programme requires 20 or more full-time US W-2 employees; it does not say whether a non-US company with that US payroll would be taken.
Its 6 rules, in its own words
- Where a holder can be based
“Our current program focuses on operating companies with at least 20 full-time U.S. W-2 employees and three years in business.”
- Where a holder can be based
“20+ full-time U.S. W-2 employees”
- Where a holder can be based
“Nyne’s current company program is for operating businesses with at least 20 full-time U.S. W-2 employees, at least three years in business, and a decision-maker authorized to approve the license.”
- Where a holder can be based
“Country USA Canada Europe Other”
- How a holder abroad is paid
“Once your dataset is delivered and accepted, you receive payment as outlined in your partnership agreement.”
- How a holder abroad is paid
“Initial payment Amount and currency for the defined package, after mandatory fees”
- Where a holder can be based
Other buyers, marketplaces and labs' programmes
AWS Data Exchange
Takes holders from a listed set of places. The US, EU member states, the UK, Switzerland, Norway, Israel, Japan, Australia, New Zealand, Hong Kong, the UAE, Bahrain and Qatar, with a W-8 and US bank details for non-US providers.
Its 8 rules, in its own words
- Where a holder can be based
“To provide data products on AWS Data Exchange, you must be a permanent resident or citizen in one of the following countries or SARs, or a business entity organized or incorporated therein: Australia¹ Bahrain¹² European Union (EU) member state¹ Hong Kong SAR Israel¹² Japan²³ New Zealand¹ Norway¹² Qatar Switzerland¹² United Arab Emirates (UAE)¹² United Kingdom (UK)¹ United States (US)”
- Where a holder can be based
“Eligible countries and regions Australia Bahrain Colombia European Union (EU) member states Hong Kong SAR India Israel Japan New Zealand Norway Qatar South Korea Switzerland United Arab Emirates (UAE) United Kingdom (UK) United States (US)”
- Where a holder can be based
“Data providers must use a valid legal entity domiciled in the United States or a member state of the EU, supply valid banking and taxation identification, and be qualified by the AWS Data Exchange business operations team.”
- Tax forms
“For US-based entities, a W-9 form and a banking account from a US-based bank are required. Non-US sellers are required to provide a W-8 form, value-added tax (VAT) or goods and services tax (GST) registration number, and US bank information.”
- Tax forms
“¹ Providers of paid products in these countries must provide VAT registration information in country of establishment.”
- How a holder abroad is paid
“AWS disburses payments monthly directly to the bank account associated with the AWS account registered as a seller, minus AWS Marketplace service fees.”
- How a holder abroad is paid
“Provide bank account information in an eligible jurisdiction that can accept USD disbursements”
- How a holder abroad is paid
“If you don't have a US bank account, you can register for a virtual US bank account from Hyperwallet .”
- Where a holder can be based
Datarade
Takes holders from anywhere. Its provider application accepts any registered business and its country list is a full world list; it excludes only sales of data to sanctioned countries and individuals or unregistered businesses.
Its 7 rules, in its own words
- Where a holder can be based
“Datarade connects data providers with qualified data buyers. Publish your products once and reach a global audience through the Datarade Marketplace.”
- Where a holder can be based
“Country Select country Afghanistan Åland Islands Albania Algeria ... Cuba ... Iran (Islamic Republic of) ... Korea (Democratic People's Republic of) ... Russian Federation ... Syrian Arab Republic”
- How a holder abroad is paid
“Commission-only Most popular For exploring the platform $0 / year + 30% commission”
- Sanctions
“Sanctioned countries: Selling Data to countries under international sanctions.”
- Where a holder can be based
“Trusted global buyers like Deloitte, Amazon, and top AI startups, Datarade is a leading B2B data marketplace offering 600+ different categories from over 500 verified, compliance-assured data providers.”
- How a holder abroad is paid
“Datarade does not host providers' data directly. The delivery and fulfilment are securely managed by the data supplier, ensuring that providers maintain control over their data.”
- Where a holder can be based
“Submit profile basics (company bio, geographic coverage, data offering, use cases supported)”
- Where a holder can be based
Google (content offer pilot)
Takes holders from anywhere. The page says it works with thousands of partners worldwide and sets no country rule, but the form behind the sign-in could not be read.
Its 3 rules, in its own words
- Where a holder can be based
“We’ve been testing this pilot for over a year with thousands of partners worldwide.”
- How a holder abroad is paid
“Partners will be paid for the delivery of their offline content, which will be used to improve our products and services, including by enhancing our quality evaluations, developing product features, and expanding opportunities for display across our products.”
- How a holder abroad is paid
“Upon invitation, we work with partners in an offer process where we mutually agree on a fee.”
- Where a holder can be based
Google (News AI pilot)
Takes holders from anywhere. It says 400+ publications worldwide take part, but it is a programme for news publishers, not for company records.
Its own words
- Where a holder can be based
“more than 400 publications worldwide are now participating in this program.”
- Where a holder can be based
Kled
Works with holders abroad. No country is required and non-US residents are expressly handled (W-8BEN), but the terms call the service US-operated and require a warranty of not being in a US-embargoed country; it pays individuals by Venmo, PayPal or Solana crypto with ID and tax checks.
Its 11 rules, in its own words
- Where a holder can be based
“The Services are operated by us in the United States. Those who choose to access the Services from locations outside the United States do so at their own initiative and are responsible for compliance with applicable local laws.”
- Where a holder can be based
“Our app has consistently remained top 100 in 20+ countries and top 10-20 in countries like Indonesia.”
- Where a holder can be based
“Define your needed dataset by modality, regions, activity, labels, and the date you need it by.”
- How a holder abroad is paid
“The Services may allow you to earn certain third-party blockchain tokens such as SOL (“ Tokens ”), fiat currency or other rewards (as determined by us in our sole discretion) for your accepted Submitted Content”
- How a holder abroad is paid
“We collect your wallet address, which is used for cryptocurrency payouts.”
- Sanctions
“By using the Services, you represent and warrant that you are not located in any such country or on any such list. You are responsible for and hereby agree to comply at your sole expense with all applicable United States export laws and regulations.”
- Tax forms
“A tax form is required when your yearly Kled earnings reach $600 or more . You must submit the form before you can cash out once this threshold is hit. Which Form Do I Need? US resident or citizen W-9 Non-US residentW-8BEN”
- Tax forms
“If your yearly earnings meet the IRS reporting threshold, Nitrility Inc. may issue a 1099-NEC or equivalent tax document.”
- Tax forms
“Tax form (if applicable) W-9 or W-8BEN required when yearly earnings reach $600+”
- How a holder abroad is paid
“Available Methods Kled supports three payout methods. You can set up one or multiple.”
- How a holder abroad is paid
“Minimum Amount You must have at least $25.00 in your available balance before you can request a cashout.”
- Where a holder can be based
Meta (Language Technology Partner Program)
Works with holders abroad. No country rule or payment is published, the one named partner is in Canada, it wants underserved-language data, and its interest form is now closed.
Its 4 rules, in its own words
- Where a holder can be based
“We are pleased that the Government of Nunavut, Canada, has agreed to work with us on this exciting initiative, collaborating to share data in the Inuit languages Inuktitut and Inuinnaqtun.”
- Language of the records
“Our efforts are especially focused on underserved languages, in support of UNESCO’s work as part of the International Decade of Indigenous Languages.”
- How a holder abroad is paid
“As a partner, you will also gain access to technical workshops led by our research teams, where you’ll learn how to leverage our open source models to build language technologies.”
- How a holder abroad is paid
“Partners will work with our teams to help integrate these languages into AI-driven speech recognition and machine translation models, which when released will be open sourced and made freely available to the community.”
- Where a holder can be based
OpenAI (Data Partnerships)
Takes holders from anywhere. Its form takes an organization of any listed country (a full world list, no exclusion shown) and the page wants data in any language, but it publishes nothing on payment or tax.
Its 3 rules, in its own words
- Where a holder can be based
“We’re already working with many partners who are eager to represent data from their country or industry. For example, we recently partnered with the Icelandic Government”
- Where a holder can be based
“Organization country * Select a country… Afghanistan Albania Algeria Andorra Angola”
- Language of the records
“We’re particularly looking for data that expresses human intention (e.g., long-form writing or conversations rather than disconnected snippets), across any language, topic, and format.”
- Where a holder can be based
Opendatabay (Opendatabay Ltd, England and Wales)
Takes holders from anywhere. Opendatabay (a UK company) says it serves users globally and takes data from any registered business that passes KYC/KYB, but it requires the holder to cover cross-border transfer in its legal basis and says nothing on tax forms or sanctions.
Its 4 rules, in its own words
- Where a holder can be based
“Opendatabay serves users globally but operates under UK law.”
- How a holder abroad is paid
“Payouts are made only to verified business accounts.”
- How a holder abroad is paid
“All payments are processed securely through Stripe and PayPal. We receive transaction IDs and user billing information for verification and record-keeping purposes. For sellers, we may store payout details such as bank account information required for withdrawals.”
- How a holder abroad is paid
“From 1 October 2026, all marketplace prices are denominated in USD. Existing listings have been converted from GBP to USD using a fixed migration rate of $1 = £0.76”
- Where a holder can be based
Protege
Takes holders from anywhere. Its pages say partner datasets represent 160 countries and name non-US partners (Cambodia, an EMEA broadcaster, South Africa), and state no eligibility by country, but publish no country rule, currency or tax form.
Its 7 rules, in its own words
- Where a holder can be based
“By combining content across 150+ global providers, we construct datasets that power multiple AI buyers and evolving model use cases.”
- Where a holder can be based
“Number of countries represented across partner datasets 160”
- Where a holder can be based
“A major EMEA-based broadcaster earned over $1M in 6 months by licensing its scripted and cinematic library with Protege”
- Where a holder can be based
“The world’s largest library of Khmer-language television from Cambodian Broadcasting Service (CBS) will power more inclusive AI models and expand Protege’s uniquely diverse, six-continent audio-visual…”
- Where a holder can be based
“We’re excited to partner with Protege to bring South Africa’s healthcare data to a global audience.”
- Language of the records
“Speaker Track Languages Covered 70+ Languages”
- How a holder abroad is paid
“Your content is included in qualified deals, and you receive transparent, timely revenue share payments.”
- Where a holder can be based
Created by Humans
Says nothing about where. It is a US-run platform that licenses books and other creative works (not company records), tells users outside the US to follow local law, and is governed by New York law with foreign law excluded.
Datacurve (Shipd)
Says nothing about where. No country is required, but contributors warrant they are not in a US-embargoed country, may face IP and VPN monitoring, and may be asked for identity and tax documents before payment; it hires contributors rather than licensing a holder's records.
Its 5 rules, in its own words
- Where a holder can be based
“Engineers from 40+ countries, architecting and solving the challenges that frontier models can't.”
- How a holder abroad is paid
“Payments for completed work are processed through Shipd's designated payment processor(s) in accordance with the rates and terms posted on the Platform or communicated to you.”
- Tax forms
“You are solely responsible for all tax obligations, withholdings, and reporting requirements arising from amounts paid to you under these Terms. Shipd may require you to provide tax documentation or certifications as a condition of payment, and you represent that any such documentation you provide is accurate and complete.”
- Sanctions
“you are not located in, a national of, or ordinarily resident in any country or territory subject to a comprehensive U.S. government embargo or economic sanctions program;”
- Sanctions
“Shipd reserves the right to implement technical measures to monitor platform access for compliance with applicable sanctions laws, including IP-based monitoring and VPN detection.”
- Where a holder can be based
DataVendor
Says nothing about where. UNREAD, the site sits behind a Vercel browser checkpoint (HTTP 429) and none of its pages could be fetched.
Defined.ai
Says nothing about where. The Data Partner page asks only for a business that generates or aggregates data and names no country condition; the Seattle company's licence and terms are governed by Washington law.
Its 2 rules, in its own words
- Sanctions
“Suppliers will not deliver datasets where its provenance is tied to countries sanctioned pursuant to EU's global human rights sanctions regime. Defined.ai Suppliers understand and comply with all applicable international trade controls (i.e., export controls, economic sanctions, and anti-boycott laws).”
- Sanctions
“The Data may be subject to US export control laws, including the Export Control Reform Act and its associated regulations. Licensee shall not, directly or indirectly, export, re-export, or release the Data to, or make the Data accessible from, any jurisdiction or country to which export, re-export, or release is prohibited by law, rule, or regulation.”
- Sanctions
Microsoft (Publisher Content Marketplace)
Says nothing about where. It names no country rule, its co-designers are US publishers, and it says PCM is for grounding and display, not training.
Its 3 rules, in its own words
- Where a holder can be based
“PCM will support publishers of all sizes, from large national and international organizations to specialized and independent voices.”
- Where a holder can be based
“Over the past several months, we have been co-designing PCM with leading U.S. publishers, including The Associated Press, Business Insider Inc, Condé Nast, Hearst Magazines, People Inc, USA TODAY Co., and Vox Media LLC, among others.”
- How a holder abroad is paid
“publishers will be paid on delivered value”
- Where a holder can be based
pre.dev (Pre Dev Corp)
Says nothing about where. Pre.dev publishes no rule on where a holder must be based, and its code-licensing sign-up takes individuals as well as companies, but no seller terms are on its public pages.
Prime Intellect
Says nothing about where. No company-records programme was found (its programme pays bounties to builders 'around the world' for open-source environments), but its site terms bar use from China, Iran, North Korea, Syria, Cuba and Crimea.
Its 4 rules, in its own words
- Where a holder can be based
“Today, we’re committing hundreds of thousands of $ as grants to sponsor builders around the world to help us grow the Environments Hub into the premier platform for open evals and RL environments.”
- Where a holder can be based
“Those persons who choose to access the Site from other locations do so on their own initiative and are solely responsible for compliance with local laws, if and to the extent local laws are applicable.”
- Sanctions
“Prime Intellect prohibits the use of its services from or to the following countries: China Iran North Korea Syria Cuba Crimea region of Ukraine.”
- Sanctions
“You represent and warrant that you are not located in, under the control of, or a national or resident of any country subject to a U.S. government embargo or designated as a state sponsor of terrorism.”
- Where a holder can be based
Sieve
Says nothing about where. Eligibility is described by what a business owns or does and by size (50 or more employees for the recording path), with no country named.
Its own words
- Language of the records
“Speech, sound, and performance Podcasts Audiobooks Multilingual recordings”
- Language of the records
Versos AI, Inc.
Says nothing about where. Versos publishes no rule on where a holder must be based (its enquiry and value forms ask no country); it works with video libraries held in the holder's own cloud storage.
Its own words
- Language of the records
“Filter and build datasets based on granular AI training specifications, including content type, format, duration, resolution, language, production status, and aspect ratio.”
- Language of the records
Wirestock
Says nothing about where. Its FAQ says anyone who creates the content and holds the licensing rights can join, with no country named, and it pays by PayPal or Payoneer; it deals in creator content (photos, illustrations, video), not company records.
Its 2 rules, in its own words
- How a holder abroad is paid
“Currently, Wirestock only supports payments through Paypal and Payoneer. Wirestock may, however, later include additional payment processors. There is a minimum payout rate per accounting period of either (i) thirty U.S. Dollars (USD $30.00) for PayPal and fifty U.S. Dollars (USD $50.00) for Payoneer;”
- Tax forms
“You are solely responsible for any payments required to be filed with or made to any tax authority with respect to your receipt of compensation under the Terms.”
- How a holder abroad is paid
Where we stand
We sell only for companies and people based in the United States, the United Kingdom and Canada for now: buyers rank those first, and each other country brings its own rules we have not taken on yet. Your valuation stays yours and binds you to nothing. The valuation works for a holder in Germany as it does anywhere, and the rest of this page is what to settle whichever way you sell.
What your records are worth, and what to settle before a buyer in the US sees them.
In this topic
Selling data
- Start hereHow do you sell or license data to AI companies?The whole path from what you hold to a signed licence: valuing it, the rights file, the round of buyers, the fee, and the special cases.
- GuideSelling Data to US AI Companies From Outside the USYes. US AI buyers license data from companies abroad, but price them lower, and your privacy law, US withholding tax and payment route change the deal.
- GuideHow to Sell Your Codebase to AI Companies (and What It's Worth)AI companies buy private codebases and price them by the tasks they yield: per repository, pull request or task. Who buys, what they pay, what kills a sale.
- GuideUS Withholding Tax When a Foreign Company Licenses DataA US buyer withholds 30 percent of a data licence fee paid abroad unless a treaty lowers it and a W-8BEN-E arrives first. Rates by country, forms and VAT.
- ToolValue your dataA valuation from public prices, in a few questions and no upload.
- PageHow prices are negotiatedThe levers that move a price, each with the deal that shows it.
- PageDirect or through a brokerWhat a holder gains and gives up by dealing with one buyer alone.
- PageAn offer in your inboxHow to read an unsolicited offer for a company's old data.
- PageClosed and insolvent companiesData is an asset of the estate, and it expires.
- PageIntroduce a holderIntroduce a data holder and earn a share of our fee.
- PageFeesWhat a holder pays us, only when a sale closes.
- RegisterSelling from your countryWhich buyers take holders from each country, its privacy law and the US tax on the fee.
- PageHow we verify holdersThe checks every holder passes before a buyer hears of its lot.
- PageToolsEvery free tool, in the order a sale uses them.