Questions about selling and buying AI data

Selling your company's data to AI companies is possible when the company owns the records, holds the rights to license them and can remove personal data it cannot lawfully disclose. Buyers such as micro1 and Mercor publish payout ranges for operational records. Most deals are licenses rather than outright sales, so you keep the original data. Book a free consultation to find out whether yours qualifies.

AI buyers want company data that records real work with outcomes: internal communications, SOPs, project histories, CRM records, support tickets, source code with its history, and professional files such as closed legal matters or tax workpapers. micro1 and Mercor both list record types like these. Public web data, contact lists on their own and data you downloaded from elsewhere rarely interest them.

Your company's data can be worth anything from a few thousand dollars to several million, depending on the industry, the size of the archive and exclusivity. ddiy.co reports a median of $280,000 per company across buyers' published offers, but only $31,600 across the few sales made public. Our free data valuation gives a realistic range and the likely buyers for your records.

A minimum number of staff matters to some buyers but not all. micro1 asks for 30 or more full-time employees, and Mercor's listings start at 5 attorneys and staff for law firms and 50 agents for call centers. Polyshares publishes no headcount minimum, according to ddiy.co. We match your size to buyers whose published criteria you meet.

An exclusive license pays more but stops you licensing the same data to anyone else. ddiy.co's 2026 pricing guide puts an exclusive deal at about 2.5 to 5 times a non-exclusive one. A non-exclusive license lets you sell to several buyers over time. Our guide to exclusive vs non-exclusive data licenses sets out the trade-offs.

You keep your data after licensing it, because a license grants the buyer rights to use a copy rather than transferring ownership of your records. Your original systems are untouched. Check the exclusivity terms, though, since an exclusive license can stop you licensing the same records to another buyer for the period it covers.

A closed or bankrupt company can sell its data if someone with authority acts for it, such as a bankruptcy trustee or the board. In a US bankruptcy, 11 USC 363(b)(1) bars a trustee from selling personally identifiable information against the debtor's privacy policy unless the sale fits the policy or a court approves it after a consumer privacy ombudsman is appointed. Google won bankrupt Spirit Airlines' archive with a $10 million bid, MIT Technology Review reported. See selling data from a closed company.

Whether your clients find out depends on what your contracts and privacy notices require you to tell them. A license can include confidentiality terms, but your notices may oblige disclosure, and some client contracts need consent before any reuse. Our guide to telling staff and clients about a data license covers when and how to do it.

Data that has trained a model cannot practically be pulled back out, according to Daniel Goldberg of Frankfurt Kurnit Klein and Selz. A license can require the buyer to delete raw copies by a set date, but the trained model keeps what it learned. Agree deletion and withdrawal terms before signing, and exclude anything you would not want learned.

Selling data through a broker or directly depends on how much time you have for buyer research and diligence. Going direct to one buyer can be quick if you meet its published criteria. A broker such as Sell My Data To AI vets buyers, prepares you for their checks and lets you compare offers. See AI data marketplaces, brokers and direct deals compared.

Source code can be sold to AI companies when you own it outright, including IP assignments from staff and contractors, and have identified any open-source components. DataVendor says labs pay more for code written mostly by people and for systems languages such as Rust, Go, C and C++. Read how to sell source code to AI companies.

Buying AI training data starts with a written requirement: the use, data type, volume, languages and license terms you need. Then shortlist sellers, review random samples under NDA, run due diligence and sign a license with acceptance tests. You can submit a data requirement at no charge, and we will find, vet and shortlist sellers for you.

Through Sell My Data To AI you can source rights-cleared data held by companies and archives, including workplace communications, support tickets, source code, agent trajectories, evaluation sets, preference data, expert domain records, speech, video, images, multilingual text, documents, call recordings and CRM data. See training data by type for what each involves.

Licensed training data comes with a contract from someone who holds the rights, while scraped data is collected from the web without one. Scraping can serve for prototyping, but production use needs provenance. In December 2023 the Stanford Internet Observatory reported more than 3,200 images of suspected child sexual abuse in the scraped LAION-5B dataset.

AI training data costs vary by type and license. Getphyllo's reported market ranges run from $5,000 to $15,000 for research licenses to $25,000 to $500,000 or more for commercial licenses, with exclusive deals at $1 million and above. Video is quoted per minute, and Bloomberg reported Adobe paying about $2.62 on average. See AI training data pricing models.

You can see a sample before buying data. Buyers see redacted samples under a mutual NDA, drawn at random across the archive rather than hand-picked, so you can judge the whole package. Full data moves only after the license is signed. Our guide on how to evaluate a dataset sample covers what to test.

You contract directly with the seller, not with Sell My Data To AI. We are a broker: we find and vet sellers, arrange samples and help agree terms, but the license is signed between you and the data owner. We do not take custody of the raw data, and we do not buy data for our own use.

Exclusive rights to a dataset can be negotiated, at a higher price. ddiy.co's 2026 pricing guide puts exclusive deals at about 2.5 to 5 times non-exclusive ones. Exclusivity matters most for evaluation sets, which lose value once shared. Agree the scope, duration and field of use of any exclusivity in the license itself.

A buyer should insist on license terms that name AI training explicitly, define what may be done with derived assets such as annotations, embeddings, fine-tuned models, weights and outputs, and allocate risk through representations, indemnities, liability caps and holdbacks. Nixon Peabody's October 2026 alert sets out these points. See our licensing agreement checklist.

A buyer's data requirement should state the intended use, data type, domain, languages, volume, date range, delivery format, license type and budget range. Add any exclusions, such as personal data or particular jurisdictions. The more specific the requirement, the easier it is to match sellers. Submit a data requirement to start.

Synthetic data can supplement licensed real data but has known gaps. MIT Technology Review reported in October 2026 that real archives avoid errors common in synthetic data, such as references to dates that do not exist, although no company has published tests comparing the two. Synthetic data generated from a licensed seed set still depends on the rights to that seed.

A data manifest is the inventory of a dataset: record counts, date ranges, source systems, languages and formats. It matters to buyers because it describes what you are paying for. Our seller vetting checks that the manifest matches what is actually delivered. Read how to build a data manifest for what a good one contains.

Sell My Data To AI vets every seller against an eight-point standard before any buyer sees their data: authority to sign, ownership and chain of title, contract restrictions, privacy notice and lawful basis, third-party content, de-identification method, representative sample and manifest accuracy. How we vet explains each check in full.

Sell My Data To AI verifies every buyer before they see a sample. We check company identity, intended use, security and retention commitments, onward transfer limits and, for larger deals, ability to pay. The seller then approves each buyer individually, so no one sees your data without your agreement.

Sell My Data To AI does not take custody of your raw data. Buyers see redacted samples under a mutual NDA, and full data moves only after you sign a license directly with the buyer. Your records stay in your control until a deal is agreed, and you can withdraw at any point before signing.

Sell My Data To AI is independent of the buyers it works with. We do not own an AI model and we do not buy data for our own use, so we have no interest in acquiring your records ourselves. Our role is to match sellers with vetted buyers and help both sides through diligence.

You approve every buyer that sees your data, as well as every sample and every license term. Nothing is shared without your agreement, and you can withdraw at any point before signing. Buyers are verified by us first, so the ones you are asked to approve have already passed our checks.

Knowing a seller really owns the data comes down to evidence of chain of title. Our vetting checks that the company created or acquired the records and holds the rights to license them, including IP assignments from staff and contractors, and that client contracts and NDAs do not prohibit the use. Read data provenance explained.

Data samples are drawn at random across the archive so a buyer can judge the whole package rather than its best rows. A hand-picked sample can hide gaps and errors. Each sample is labeled with its sampling method, which lets the buyer check it against the data manifest.

Proper de-identification is checked in two steps. The seller removes direct and indirect identifiers by a documented method, and a re-identification test is run on a sample. Both are part of our seller vetting. HIPAA's Expert Determination method in 45 CFR 164.514(b) sets a US benchmark: a qualified expert must find that the risk of identifying anyone is very small. For UK data, the ICO's motivated intruder test asks whether a determined person could identify individuals.

What stops a buyer from leaking or reselling your data is vetting plus contract. We check each buyer's security and retention commitments and onward transfer limits before they see a sample, samples are shared under a mutual NDA, and the license you sign sets the rules on onward transfer, deletion and remedies.

Selling company data to AI can be done safely when you exclude high-risk records, de-identify the rest by a tested method and limit what the buyer may do with it. The main risks are privacy claims, breaches of client contracts and data that cannot be withdrawn from a trained model. Read is selling company data to AI safe?

The vetting file buyers receive covers each shortlisted seller: rights evidence, the provenance record, the de-identification method, the data manifest and the proposed license terms. It lets your legal and data teams run their own diligence on top of ours before you agree to review a sample.

Selling company data for AI training is legal in principle, but whether a particular sale is lawful depends on what the data contains, the contracts and privacy notices that govern it and the laws that apply, such as the CCPA, HIPAA or GLBA in the US. Holding data does not mean you have the right to sell it. We are a broker, not a law firm, so take independent legal advice.

In the US no single federal privacy law applies. California's CCPA and the other comprehensive state privacy laws cover consumer data, HIPAA covers health records, GLBA covers financial institutions' customer data and the FTC acts against broken privacy promises. If you hold UK or EU data, UK GDPR or EU GDPR also applies and requires a lawful basis or proper anonymization. Read UK GDPR, EU GDPR and CCPA when selling data.

Pseudonymized data is not deidentified under California's CCPA unless the business also meets the three conditions in Civil Code section 1798.140(m): reasonable measures, a public commitment not to re-identify and contracts binding recipients. The CCPA defines pseudonymization separately, in section 1798.140(aa). If you hold UK data, pseudonymized data is still personal data under UK GDPR. Buyers should ask which standard the data meets. See pseudonymization.

De-identification is the term US law uses. HIPAA sets two de-identification methods in 45 CFR 164.514(b), and the CCPA sets its own deidentified standard in section 1798.140(m). Anonymization is the UK and EU term for the outcome where no one can reasonably identify a person, which takes data outside UK GDPR. FTC staff say data is anonymous only when it can never be associated back to a person. See de-identification.

Data covered by an NDA can only be licensed if the NDA's terms allow it or the other party consents. Daniel Goldberg of Frankfurt Kurnit Klein and Selz notes that email and chat often contain other companies' confidential information under NDAs and commercial agreements. Our vetting checks contract restrictions, and material you cannot clear should be removed.

Your privacy notice matters because you must honor the notice in force when the data was collected. FTC staff warned in February 2024 that adopting more permissive data practices, such as using data for AI training, through a surreptitious, retroactive amendment to a privacy policy may be unfair or deceptive. If the notice did not permit the disclosure, exclude the personal data or de-identify it properly before any sale.

A medical practice will usually need to de-identify patient data before licensing it for AI training. HIPAA allows de-identification by expert determination or by removing 18 identifiers under 45 CFR 164.514(b), and HHS says de-identified health information is no longer protected health information. If you hold UK patient data, health data is special category data under UK GDPR. See our page for medical practices.

Staff messages can be included in a data sale only after checking employment law and privacy rules. In the US, messages about pay and working conditions may be protected concerted activity under Section 7 of the National Labor Relations Act, and California's CCPA has covered employee personal information since January 1, 2023. In the UK they are personal data under UK GDPR. Suppliers interviewed by MIT Technology Review exclude HR records altogether.

AI developers in California must disclose some training data details. AB 2013 requires developers of generative AI systems to post a high-level summary of the datasets used in development on their websites, on or before January 1, 2026, according to the California Attorney General. In the EU, Article 53(1)(d) of the EU AI Act requires providers of general-purpose AI models to publish a summary of training content. See the EU AI Act and training data transparency.

The CCPA, at California Civil Code section 1798.140(ad), defines a sale as making a consumer's personal information available to a third party for monetary or other valuable consideration, so a paid data license that includes personal information can count. Deidentified information is excluded from personal information, and the deidentified standard in section 1798.140(m) requires reasonable measures, a public commitment not to re-identify and contractual obligations on recipients.

You should take independent legal advice before signing any data license. We are a broker, not a law firm: we tell you what buyers will check and help you prepare, but a lawyer should review the license, your contracts and your privacy position for your own case before you sign.

Sell My Data To AI charges sellers nothing to be assessed. We charge a success fee only when a deal completes, and the fee is agreed in writing before we introduce you to any buyer. If no deal completes, you pay nothing. See fees for how it works.

Sell My Data To AI charges buyers nothing to submit a requirement or review a shortlist. Any fee is agreed in writing before an introduction, so you know the cost before you speak to a seller. Submit a data requirement or book a free consultation to start.

The consultation is free and carries no obligation. It is a 30-minute video or phone call for data owners and AI teams. Book a free consultation if you hold data, or book as a buyer if you are sourcing it. Please do not send any data before or during the call.

If a deal does not complete, a seller pays us nothing, because our success fee applies only when a deal completes. You can withdraw at any point before signing, and you approve every buyer, sample and license term along the way, so nothing is committed until you sign.

The data valuation is free, because sellers pay nothing to be assessed. It gives you a view of likely buyers, a realistic range and the rights issues to clear before a sale. There is no obligation to go ahead afterward. Request a data valuation to start.

Sellers get paid by the buyer under the payment terms in the license they sign together. Published timings vary: according to ddiy.co, Mercor pays 2 to 4 weeks after signing and Polyshares works on net 30 to 60 days, while micro1's own page says it pays within 30 days of invoicing after a review period of up to 30 days.

Data license payments can be made upfront, in milestones or as a revenue share. ddiy.co reports that Handshake AI pays by milestones and that Labelbox Prism combines an upfront fee with a revenue share. Staged payments tied to acceptance tests protect a buyer if the delivered data differs from the sample.

A success fee is a fee paid to a broker only when a deal completes. Sell My Data To AI works on a success fee for sellers, agreed in writing before we introduce you to any buyer, so you pay nothing if no license is signed. The glossary entry on the success fee has more.

Fees to sellers differ by buyer and platform. According to ddiy.co, Mercor and Polyshares state no fee to sellers, while FileYield's FAQ lists 8% for sellers and Opendatabay has been reported to charge tiered commission of 5 to 30% by sale price. Compare the net amount you will receive, not only the headline offer.

A revenue share and a flat fee suit different sellers. A flat fee gives certainty and is paid on a known schedule. A revenue share can pay more over time if the buyer licenses widely, but depends on the buyer's sales and reporting. Troveo, for example, takes no fees and is paid from a revenue share, according to ddiy.co.

Who pays for preparing and de-identifying the data depends on the buyer. According to ddiy.co, Mercor covers extraction and Scale AI anonymizes inside the seller's environment. Elsewhere the seller prepares the export. Agree who bears preparation costs in writing before work starts, since de-identification of large archives takes real effort.

The process for sellers runs from a free consultation to a signed license. We assess your data and the rights to it, run our eight-point vetting, prepare a redacted sample and manifest with you, introduce verified buyers you approve and help you agree terms. You then sign the license directly with the buyer. See how it works.

The process for buyers starts with a requirement. We find and vet sellers, send you a shortlist with a vetting file for each, arrange redacted samples under a mutual NDA and help agree acceptance tests, staged payments and remedies. You then contract directly with the seller. Submit a data requirement to begin.

A data deal's length depends on the buyer and on how ready your data is. Published timings include an offer in 1 to 2 weeks from Polyshares and a 7 to 14 business day process from Handshake AI, according to ddiy.co. Rights checks, de-identification and buyer review can add time on top of the commercial steps.

The first consultation is a 30-minute video or phone call. You tell us what records you hold or what data you need, and we tell you honestly whether a deal makes sense and what it would take. You do not need to prepare anything detailed, and you should not send any data before or during the call.

We reply to inquiries within two business days. You can reach us through the contact page or by email at hello@sellmydatatoai.net. Describe the records you hold or the data you need in general terms, and do not attach any data to your message.

Before selling data you need a rough inventory: which systems hold the records, how many years they cover, approximate volumes and languages. Later you will need a data manifest, a representative sample and evidence of ownership. Read how to build a data manifest for the detail.

Data is delivered to the buyer only after you sign a license, as a fixed export rather than access to your systems. Frankfurt Kurnit Klein and Selz advises against giving buyer tools direct access to company systems and recommends a secure transfer method. We do not take custody of the raw data at any stage.

You can withdraw at any point before signing a license. You approve every buyer, every sample and every license term, so nothing goes ahead without your agreement, and a seller pays no fee unless a deal completes. Once a license is signed, its own terms govern any termination.

A staged or pilot data deal starts with a limited sample in a controlled environment, then expands access as agreed conditions are met. Nixon Peabody recommends this approach in its October 2026 alert on buying training data. A pilot lets a buyer test quality and rights before paying for the full archive.

A mutual NDA binds both seller and buyer to keep each other's information confidential. It is signed before a buyer sees any redacted sample, so the sample and the discussions around it are protected from the start. The glossary entry on the mutual NDA has more.

After the license is signed, the full data moves from seller to buyer under the agreed delivery and security terms, the buyer runs its acceptance tests and payments follow the schedule in the license. For sellers, our success fee becomes due only when the deal completes, on the terms agreed in writing before the introduction.