# 10 Best Platforms for Selling Data to AI Companies (2026)

Source: https://sellyouraidata.com/guides/platforms-for-selling-data-to-ai-companies/
Updated: 2026-10-26

> As of October 2026, the main platforms for selling data to AI companies are SellYourAIData, micro1, Handshake AI, Scale AI, Mercor, Turing's Project Lazarus, SimpleClosure Asset Hub, Protege, Troveo and Korra. They fall into three models: buyer programs that purchase or license your records directly, marketplaces and brokers that match you with buyers, and seller-side representatives that run several buyers against each other. Published payouts range from $10,000 to more than $4 million, but every figure is an indicative anchor set by one buyer, so the platform model and the contract terms matter more than the headline number.

## Key takeaways

- There are three ways to sell business data to AI companies: a buyer's own program (micro1, Handshake AI, Scale AI, Mercor, Turing), a marketplace or broker (SimpleClosure, Protege, Troveo, Korra), or a seller-side representative such as SellYourAIData that brings several buyers to one deal.
- Published payout ranges run from $10K to $1M+ at Scale AI, $100K to $1M+ at micro1, $100K to $4M+ at Handshake AI and up to $1M at Turing. Each company describes its figure as illustrative or indicative, not an offer.
- Documented completed sales are rarer than advertised ranges. Shut-down startups reported $10,000 to $100,000 per deal through SimpleClosure, and Google's winning bid for Spirit Airlines' data was $10 million.
- The data types buyers name most often are workplace chat and email, documents and procedures, tickets and project records, CRM records and source code.
- Who pays the platform matters: Korra and Mercor say the seller pays nothing, Datarade charges providers 15% to 30% commission plus optional subscriptions, and seller-side representatives are paid from a completed deal.
- No public program publishes a full contract. Exclusivity, field of use, term, ownership, de-identification liability and payment timing have to be settled in writing before any price means anything.

## The short answer

The best platforms for selling data to AI companies in 2026 are, in our ranking: **SellYourAIData** (seller-side representation with competing buyers), **micro1** (highest published floor for operational data), **Handshake AI** (widest published range), **Scale AI** (recurring partnerships), **Mercor** (connector-based, costs covered), **Turing's Project Lazarus** (single payment for engineering and decision history), **SimpleClosure Asset Hub** (companies that are shutting down), **Protege** (healthcare and media), **Troveo** (video and business data) and **Korra** (a marketplace where you choose among offers).

These platforms work in three different ways, and the model decides more about your outcome than the headline number does.

1. **Buyer programs.** The company that wants your data runs the program and sets the price: micro1, Handshake AI, Scale AI, Mercor and Turing.[^micro1-data][^hs-license][^scale-dp][^mercor-data][^lazarus]
2. **Marketplaces and brokers.** A third party lists your data or introduces you to buyers it chooses: SimpleClosure, Protege, Troveo, Korra and others.[^sc-assethub][^alleywatch][^troveo][^korra]
3. **Seller-side representation.** A firm works for you, not the buyer, and runs several buyers against each other: SellYourAIData.

If you only read one more section, read [how to choose](#how-to-choose-the-right-platform), and if you already have an offer on the table, get a [free offer review](/offer-review/) before you sign.

## How we ranked these platforms

We looked at every company we could find that publicly invites businesses to sell or license data to AI developers, using the research behind our [AI data buyers guide](/guides/ai-data-buyers/) and our [deal tracker](/deals/). We opened each company's own program page in October 2026 and cross-checked figures against news coverage. Ten made this list. The rest appear under [other platforms](#other-platforms-worth-knowing) or [companies that are not data buyers](#companies-that-are-not-data-buyers).

We scored each platform on six questions a seller should ask:

| Criterion | What we checked | Why it matters |
|---|---|---|
| Price transparency | Is any payout figure published, and is it labeled an offer or an estimate? | A published anchor lets you test an offer; a blank leaves the number to the buyer |
| Competition | Does the seller see more than one buyer's offer? | One bidder sets the price; several bidders discover it |
| Who pays the fee | Does the seller, the buyer or nobody pay the platform? | Fees come out of your net proceeds |
| De-identification | Who removes personal information, and who carries the risk if it fails? | This is the largest legal exposure in a data sale |
| Terms disclosed | Are exclusivity, ownership, field of use and term stated? | Terms move value more than price does |
| Track record | Is there a named, documented completed deal? | Programs are new; most claims are self-reported |

> **Note:** SellYourAIData publishes this guide and ranks itself first. We explain why in [its section](#sellyouraidata), and every other entry is limited to what the company publishes or what reputable press has reported, with a numbered source. We are not affiliated with any other company on this list.

## All 10 platforms compared

| # | Platform | Model | Best for | Published payout | Seller pays a fee? | Who de-identifies | Exclusivity stated? |
|---|---|---|---|---|---|---|---|
| 1 | **SellYourAIData** | Seller-side representative | Operating companies that want competing offers | Our working range: $50K to $500K for 50 to 200 staff | From a completed deal only | Seller-controlled sample, then buyer | Negotiated per deal |
| 2 | **micro1** | Buyer program | Large operational datasets | $100K+, $500K+, $1M+ tiers[^micro1-data] | Not stated | micro1[^micro1-data] | Not stated |
| 3 | **Handshake AI** | Buyer program | Documents and workflows across 11 sectors | $100K to $4M+, indicative[^hs-license] | Not stated | Handshake[^hs-blog] | Not stated |
| 4 | **Scale AI** | Buyer program | Recurring warehouse and work-tool data | $10K to $1M+, illustrative[^scale-dp] | Not stated | Inside your environment[^scale-dp] | Not stated |
| 5 | **Mercor** | Buyer program | Data in connected SaaS tools | Not published[^mercor-data] | No fees, says Mercor[^mercor-data] | Mercor, automated[^mercor-data] | Not stated |
| 6 | **Turing (Lazarus)** | Buyer program | Engineering and decision history, single payout | Up to $1M[^lazarus] | Not stated | Turing[^lazarus] | "Clean rights transfer"[^lazarus] |
| 7 | **SimpleClosure Asset Hub** | Wind-down broker | Companies that are shutting down | $10K to $100K, reported[^gizmodo] | Free for dissolution customers[^sc-webull] | SimpleClosure[^sc-assethub] | Not stated |
| 8 | **Protege** | Licensing platform | Healthcare and media rights holders | Revenue share, % not stated[^pulse2-a1] | Not stated | Protege, where appropriate[^alleywatch] | Not stated |
| 9 | **Troveo** | Licensing platform | Video, and business data from 30+ systems | Estimate on request[^troveo] | Not stated | Not stated | Not stated |
| 10 | **Korra** | Marketplace | Sellers who will clean and deliver the data themselves | Offers from buyers[^korra] | No, buyer pays Korra[^korra] | The seller[^korra] | License set between you and buyer[^korra] |

{{chart:payouts}}

> **For data sellers:** Every bar above except the reported ones is a single buyer's own anchor. None shows what a second buyer would pay for the same records. The one large documented auction, Spirit Airlines, closed at $10 million after a $5 million opening bid, with other bidders in the room.[^time-spirit] Competition, not the program page, set that price. See [what Spirit Airlines teaches sellers](/blog/spirit-airlines-email-sale/).

## The 10 best platforms for selling data to AI companies

### 1. SellYourAIData {#sellyouraidata}

**Model:** seller-side representation. **Best for:** operating companies with several years of business records that want more than one offer and terms written for the seller.

SellYourAIData does not buy data. It works for the company selling it. We estimate what your data is worth, write a summary without your company name, and then bring it to several qualified buyers under NDA so that offers compete. We negotiate the terms that the buyer programs leave blank: exclusivity, field of use, term, de-identification liability, audit and deletion rights, and payment timing.

- **What it takes:** business records such as [support conversations](/data/customer-service/), [internal communications](/data/communications/), [documents and procedures](/data/documents/), [workflows and tickets](/data/workflows/), [CRM history](/data/crm-history/), [source code](/data/code/) or [healthcare data](/data/healthcare/) where the rights allow it.
- **Payout:** depends on the data and the bidders. Our working range for a company of 50 to 200 employees is $50,000 to $500,000 per license. That is our estimate, not a guarantee. Get a range for your own data from the [free valuation estimator](/valuation/), and see the evidence behind it on [market rates](/market-rates/).
- **Fees:** no upfront fee. Our fee comes from a completed deal.
- **Privacy:** SellYourAIData does not de-identify or alter data. The owner removes personal information before anything is shared, under a mutual NDA. See [trust and security](/trust/).
- **Exclusivity:** non-exclusive by default, with exclusivity priced separately if you want it.
- **Track record:** we launched in 2026 and have no completed deals to report yet. We track the public ones on our [deal tracker](/deals/).

**Why it ranks first:** it is the only model on this list where the platform is paid by the seller to get the seller a higher price, and where several buyers bid on the same records. Every buyer program below sets its own price, and most marketplaces leave the terms to you. Start with the [three-minute assessment](/assessment/), read [how it works](/how-it-works/), or, if you already have an offer, request a [free offer review](/offer-review/).

**Watch out:** representation adds a fee and a step. If you want the fastest single payment and are comfortable with one buyer's price, a direct program can be quicker.

### 2. micro1 {#micro1}

**Model:** buyer program. **Best for:** companies with large, well-documented operational datasets in software, finance, support, sales, legal or operations.

micro1 announced on October 9, 2026 that it would spend $1 billion over 12 months acquiring and licensing enterprise operational data, financed by Citi and Hercules Capital, to build reinforcement learning environments that mirror real business work.[^micro1-blog][^tnw-micro1] Its program page lists standard operating procedures, knowledge bases, internal documentation, CRM data, project histories, QA processes, decision patterns and human feedback on AI outputs.[^micro1-data]

- **Payout:** three published tiers: $100,000 or more for qualified enterprise partnerships, $500,000 or more for large-scale or ongoing datasets, and $1 million or more for highly unique data. Compensation depends on size, workflow complexity, quality, domain expertise and uniqueness.[^micro1-data]
- **Process:** assessment, evaluation and a discovery call, then an agreement that defines scope, confidentiality and redaction. Personal information is removed where relevant, some data may be rewritten synthetically, and companies can review samples first.[^micro1-data]
- **Ownership:** micro1 says companies keep ownership of their underlying data, with deletion at the end of the engagement under the agreed terms.[^micro1-data]
- **Track record:** TechCrunch reported a $500 million gross annual run rate in August 2026, citing a person familiar with the company.[^tc-micro1] Time and CSO Online reported a late $12.5 million offer for Spirit Airlines' data.[^time-spirit][^cso] We found no named company paid under the data program.

**Watch out:** the $1 billion is a spending ceiling, not a count of deals, and the program is weeks old. Payment timing and exclusivity are not published. More in our analysis of [micro1's $1 billion commitment](/blog/micro1-1-billion-company-data/) and [Mercor vs micro1](/compare/mercor-vs-micro1/).

### 3. Handshake AI {#handshake-ai}

**Model:** buyer program. **Best for:** companies whose value sits in financial models, process documentation, technical documents and production records across a wide range of industries.

Handshake's data partnership page lists eleven sectors, including finance and insurance, hardware and engineering, healthcare, legal, logistics, manufacturing, pharmaceuticals, professional services, real estate, retail and software.[^hs-dp] Its license page names sources from AutoCAD and NetSuite to QuickBooks, ServiceTitan and Slack.[^hs-license] Handshake says it purchases data directly from companies to build environments where AI models learn to do real tasks.[^hs-blog]

- **Payout:** qualified partnerships typically range from $100K to $4M or more.[^hs-blog] The estimate form says it is "an indicative estimate only and is not an offer to purchase your data."[^hs-license]
- **Process:** each engagement is under an NDA, data is scoped and de-identified, and you can review who may receive it and how it may be used.[^hs-license][^hs-dp]
- **Track record:** TIME reported the CEO's claim that the AI unit passed $1 billion in gross annualized revenue, which concerns its expert-labor business, not data purchases.[^time-lord]

**Watch out:** final payout depends on due diligence and acceptance of terms that are not published, and the buyers are unnamed.[^hs-license] Business Insider reported that some contractors said Handshake withheld pay, which concerns expert work, not data partnerships.[^15] Compare in [Handshake AI vs Mercor](/compare/handshake-ai-vs-mercor/).

### 4. Scale AI {#scale-ai}

**Model:** buyer program. **Best for:** established companies with data in warehouses and enterprise systems that want a recurring arrangement rather than a one-time sale.

Scale's data partnership page asks for de-identified records from data warehouses (BigQuery, Snowflake, Databricks), chat tools, email and calendar, documents, work tools such as Jira, Confluence and Notion, and business platforms such as Salesforce, Workday, ServiceNow, SAP and Zendesk.[^scale-dp]

- **Payout:** an illustrative $10K to $1M or more per partnership, scaling with cadence, designed for recurring revenue.[^scale-dp]
- **Process:** Scale says it anonymizes everything inside your environment before extraction, with per-package approval, a full audit trail and on-premises or VPC options.[^scale-dp]
- **Track record:** Meta took a 49% stake at a $29 billion valuation in June 2025.[^tc-meta] Scale appointed Francis deSouza as CEO effective August 2026.[^scale-ceo]

**Watch out:** after the Meta deal, OpenAI said it was winding down work with Scale.[^tc-openai-scale] Ask who the downstream recipients are and whether any competes with you. License scope and exclusivity are not stated. See [Scale AI vs Surge AI](/compare/scale-ai-vs-surge-ai/) and [Invisible vs Scale AI](/compare/invisible-vs-scale-ai/).

### 5. Mercor {#mercor}

**Model:** buyer program. **Best for:** companies whose records live in connected SaaS tools and that want the extraction work done for them.

Mercor's data page lists Slack, Gmail, Salesforce, GitHub and Notion among more than 100 supported tools, plus industry records such as contracts, clinical notes, underwriting files and work orders. It describes the buyers only as AI labs.[^mercor-data]

- **Payout:** no amounts published. Pay depends on volume, the tools connected and depth of data.[^mercor-data]
- **Fees:** Mercor says there are no fees or deductions and that it covers extraction, anonymization and transfer.[^mercor-data]
- **Payment:** by wire when the anonymized dataset is delivered to the buyer, typically two to four weeks after signing.[^mercor-data]
- **Track record:** raised $350 million at a $10 billion valuation in October 2025,[^tc-mercor-c] and bid $7.5 million in the Spirit Airlines auction.[^time-spirit]

**Watch out:** Mercor disclosed a 2026 security incident affecting its expert platform,[^mercor-incident] and TechCrunch reported that Meta paused contracts.[^tc-mercor-month] Ownership and license terms are not stated. See [Mercor vs Turing](/compare/mercor-vs-turing/).

### 6. Turing (Project Lazarus) {#turing-project-lazarus}

**Model:** buyer program. **Best for:** software companies, often winding down, with rich engineering history and records of how decisions were made.

Project Lazarus asks for communication threads where decisions were made, source code with reviews and tickets, wikis, runbooks and schemas, and the reasoning behind decisions.[^lazarus]

- **Payout:** a single payment of up to $1 million, confirmed after sampling, usually within a week.[^lazarus]
- **Process:** NDA, sample review, then one valuation and one contract with no recurring commitment. Turing says it works from a copy, removes personal data and secrets, and holds SOC 2 Type II.[^lazarus]
- **Track record:** raised $111 million at a $2.2 billion valuation in March 2025.[^tc-turing] We found no named Lazarus seller.

**Watch out:** "clean rights transfer" and a single payment suggest a sale, not a license. If so, you could not license the same records to anyone else. The page's pricing section displayed "Up to $0.00M" when we read it.[^lazarus]

### 7. SimpleClosure Asset Hub {#simpleclosure-asset-hub}

**Model:** wind-down broker. **Best for:** startups that are shutting down and want the data sale handled with the dissolution.

Asset Hub sells source code and workspace data from Slack, Drive, email, Notion and Figma to buyers it describes as AI foundation labs, RL environment builders and other technology companies, and says every dataset is processed for personal information before licensing.[^sc-assethub]

- **Payout:** Forbes, relayed by Gizmodo, reported $10,000 to $100,000 per company across nearly 100 deals in a year.[^gizmodo] MIT Technology Review reported more than 150 deals.[^mit]
- **Fees:** free for dissolution customers; fees for others are not stated.[^sc-webull]

**Watch out:** these prices come from closing companies with little leverage. MIT reported privacy objections, including a lawyer who called such data "legally radioactive."[^mit] Read [shutdown data sales](/blog/shutdown-data-sales/) and [legally radioactive data](/blog/legally-radioactive-data-sale-risks/).

### 8. Protege {#protege}

**Model:** licensing platform. **Best for:** healthcare organizations and media rights holders with clinical records, imaging, video, audio or motion capture.

Protege connects private data holders with AI developers, and its CEO described licensing medical imaging, clinical notes, patient histories, video, audio and motion capture.[^pulse2-protege] It raised a $30 million Series A1 led by Andreessen Horowitz in January 2026, for about $65 million in total.[^alleywatch]

- **Payout:** not published. Coverage of the round says partners get a revenue share when their data is used, with no percentage stated.[^pulse2-a1]
- **De-identification:** offered where appropriate, as part of curation.[^alleywatch]

**Watch out:** a revenue share with no stated rate cannot be priced in advance, and income depends on how often buyers use the data. For HIPAA-covered data, see our [healthcare data page](/data/healthcare/).

### 9. Troveo {#troveo}

**Model:** licensing platform. **Best for:** video and audio libraries, and companies curious about an estimate for business data from 30+ systems of record.

Troveo lists business, video, audio, gaming, text and robotics data, and invites companies to turn email, chat, CRM, tickets and files into recurring revenue.[^troveo] It claims 7,000+ data providers and 350+ companies, both self-reported.[^troveo]

- **Payout:** not published; you request an estimate.[^troveo] Reporting in 2025 said OpenAI, Google and others paid creators $1 to $4 per minute for unpublished video through intermediaries including Troveo.[^decoder-video]

**Watch out:** the documented track record is in video. We found no public report of a business paid for email, chat or CRM data through Troveo.

### 10. Korra {#korra}

**Model:** marketplace. **Best for:** sellers who can clean and deliver their own records and want to see several offers.

Korra lets you describe your records, then approved buyers make offers. It says buyers want records that show work from start to finish, such as Slack threads, Jira tickets, support cases and sales notes, and advises leaving out direct messages, HR records, legal advice and health details.[^korra]

- **Fees:** you pay nothing; the buyer pays Korra.[^korra]
- **Control:** you see every offer and approve one or none; the license is between you and the buyer.[^korra]

**Watch out:** Korra says it does not take, open, clean or deliver records, so de-identification and delivery are your job.[^korra]

## How to choose the right platform

Match the platform to your situation, not to the biggest number on a landing page.

| Your situation | Platforms to look at first | Why |
|---|---|---|
| Operating company, wants the best price and terms | [SellYourAIData](#sellyouraidata), then direct programs as comparison bids | Competing offers and seller-side terms |
| Shutting down in the next few months | [SimpleClosure](#simpleclosure-asset-hub), [Turing](#turing-project-lazarus), SellYourAIData | Speed and a single payment, but leverage is low |
| Large operational archive, 200+ staff | [micro1](#micro1), [Handshake AI](#handshake-ai), [Scale AI](#scale-ai), with representation | Highest published floors |
| Healthcare records or imaging | [Protege](#protege), SellYourAIData | HIPAA de-identification needs specialist handling |
| Video, audio or creative library | [Troveo](#troveo), Human Native, Defined.ai | Per-minute and per-asset markets exist |
| Can de-identify and deliver in-house | [Korra](#korra), cloud exchanges | Lower fees, more of the work on you |
| Already holding an offer | [Free offer review](/offer-review/) | Check price and terms before signing |

### What buyers say they want, by data type

| Data type | Named by | Our page |
|---|---|---|
| Workplace chat and email | Mercor, Scale AI, Handshake AI, Turing, SimpleClosure, Korra, AI26[^mercor-data][^scale-dp][^hs-license][^lazarus][^sc-assethub][^korra][^ai26] | [Communications](/data/communications/) |
| Documents, SOPs and knowledge bases | micro1, Handshake AI, Turing, Scale AI[^micro1-data][^hs-dp][^lazarus][^scale-dp] | [Documents](/data/documents/) |
| Tickets, projects and workflows | Scale AI, Turing, Korra, micro1[^scale-dp][^lazarus][^korra][^micro1-data] | [Workflows](/data/workflows/) |
| CRM records | micro1, Mercor, Scale AI[^micro1-data][^mercor-data][^scale-dp] | [CRM history](/data/crm-history/) |
| Source code and engineering history | Mercor, Turing, SimpleClosure[^mercor-data][^lazarus][^sc-assethub] | [Code](/data/code/) |
| Support conversations | Scale AI, Korra, Troveo[^scale-dp][^korra][^troveo] | [Customer service](/data/customer-service/) |
| Clinical notes and imaging | Protege, Mercor[^pulse2-protege][^mercor-data] | [Healthcare](/data/healthcare/) |
| Marketing performance data | None found | [Marketing](/data/marketing/) |

For why some types command more than others, see [which enterprise data types AI developers value](/research/enterprise-data-types-ai-value/).

## What it costs to sell: fees and net proceeds

Fees flow in different directions depending on the model. Mercor and Korra say the seller pays nothing,[^mercor-data][^korra] SimpleClosure is free for its dissolution customers,[^sc-webull] and on a general data marketplace such as Datarade the seller pays both a subscription and a commission.[^datarade]

{{chart:datarade}}

A platform that is paid by the buyer is not free to you: its incentive is to close the deal at a price the buyer accepts. A seller-side representative is paid by you, so its incentive is to raise the price and tighten the terms. Our [direct vs broker guide](/guides/selling-direct-vs-broker/) works through the net-proceeds math, and [pricing models in AI data licensing](/research/ai-data-pricing-models/) covers one-time, recurring and revenue-share structures.

## Before you sell: a five-step checklist

1. **Inventory what you hold.** List systems, years covered, volume and which records contain personal or customer data. The [assessment](/assessment/) does this in three minutes.
2. **Check your rights.** Read customer contracts, privacy policies and employee notices for limits on reuse. See [legal frameworks for selling business data](/research/legal-frameworks-selling-business-data/).
3. **Plan de-identification.** Decide who removes personal information, to what standard, and who carries the risk if it fails. HIPAA, GDPR and the UK ICO each set a different bar.[^ecfr][^gdpr26][^20]
4. **Get more than one price.** Use the [valuation estimator](/valuation/) and [market rates](/market-rates/) as a reference, then put the data in front of several buyers.
5. **Negotiate terms before price.** Exclusivity, field of use, term, ownership of derived datasets, audit and deletion, warranties and payment timing. The FTC has warned companies that quietly changing privacy terms to allow AI training can be unfair or deceptive.[^ftc2]

> **Watch out:** no platform on this list publishes a full contract. A blank on a program page is not a negative, it is a term you still have to negotiate in writing.

## Other platforms worth knowing

These did not make the top ten because their focus is narrower, their terms are thinner, or they are listing venues rather than buyers.

| Platform | What it is | Key fact |
|---|---|---|
| AI26 | Broker for historical company records | "Upto $500k in payouts" but no fixed payout; one-time license, about 60 days[^ai26] |
| Defined.ai | AI training data marketplace | Reuters reported 2024 prices of $1 to $2 per image and $0.001 per word charged to buyers[^decoder-spree] |
| Datarade | General data marketplace | Providers pay 15% to 30% commission by tier[^datarade] |
| AWS Data Exchange | Cloud data exchange | 3,500+ datasets from 300+ providers; not AI-specific[^awsdx] |
| Snowflake and Databricks Marketplaces | Cloud data exchanges | Provider-set listings and private exchanges[^snowflake][^databricks] |
| Human Native | Content licensing marketplace | Acquired by Cloudflare in January 2026; creators and publishers[^human-native] |
| Petrarch | Data reseller | Y Combinator Summer 2026; sources codebases and project files[^petrarch] |
| Sunset | Wind-down broker | Values data by headcount, years and volume, per MIT[^mit] |

## Companies that are not data buyers

Surge AI, Appen, Toloka, Snorkel AI, Labelbox and Invisible are among the best known names in AI data, but their own pages describe building new data with paid experts, not buying your records.[^surge][^appen-ots][^toloka-about][^snorkel][^labelbox][^invisible] If a broker tells you one of them will buy your archive, ask for evidence. Our comparisons cover them in detail: [Appen vs Toloka](/compare/appen-vs-toloka/), [Snorkel AI vs Labelbox](/compare/snorkel-ai-vs-labelbox/) and [Scale AI vs Labelbox](/compare/scale-ai-vs-labelbox/).

AI labs also buy, but mostly from large rights holders. Reuters reported Reddit's licensing deal with Google at about $60 million a year,[^reddit-reuters] and Google won the Spirit Airlines data auction with a $10 million bid.[^time-spirit] We found no lab that publishes an intake page for ordinary businesses. See the [Reddit and Google deal](/deals/reddit-google-licensing/) and the [Spirit Airlines deal](/deals/spirit-airlines-google/) on our tracker, and the wider picture in [the state of business data licensing, 2026](/research/state-of-business-data-licensing-2026/) and [reported AI data licensing deals, 2023 to 2026](/research/ai-data-licensing-deals-review/).

## Bottom line

The market for business data is real, and in 2026 more companies than ever publish a way to sell to them. But almost every published number is one buyer's anchor, almost no contract is public, and the documented sales mostly come from companies with little leverage. The platform that gets you the best result is the one that puts several buyers in front of your data and writes the terms for you. For the full picture, read [the ultimate guide to selling data to AI companies](/guides/selling-data-to-ai-companies/), browse the [marketplace](/marketplace/), or [get a free valuation](/valuation/).

[^micro1-data]: [Turn existing workflows into revenue](https://www.micro1.ai/data) (micro1, accessed October 2026)
[^micro1-blog]: [micro1 commits $1B to enterprise data acquisition](https://www.micro1.ai/blog/micro1-commits-1b-to-enterprise-data-acquisition) (micro1, October 9, 2026)
[^tnw-micro1]: [micro1 commits $1bn to buy company data for training AI agents](https://thenextweb.com/news/micro1-1bn-enterprise-data-ai-training-citi-hercules) (The Next Web, October 2026)
[^tc-micro1]: [AI data startup Micro1 reaches $500M gross run rate amid AI training boom](https://techcrunch.com/2026/08/20/ai-data-startup-micro1-reaches-500m-gross-run-rate-amid-ai-training-boom/) (TechCrunch, August 2026)
[^cso]: [Bidding war for defunct Spirit Airlines' employee data will not die](https://www.csoonline.com/article/4218769/bidding-war-for-defunct-spirit-airlines-employee-data-will-not-die.html) (CSO Online, September 2026)
[^hs-license]: [Handshake AI data licensing: get started](https://joinhandshake.com/ai/data-partnerships/license/) (Handshake, accessed October 2026)
[^hs-blog]: [Introducing Handshake AI employer data partnerships](https://joinhandshake.com/blog/our-team/handshake-ai-employer-data-program/) (Handshake blog, 2026)
[^hs-dp]: [Handshake AI data partnerships](https://joinhandshake.com/ai/data-partnerships/) (Handshake, page dated October 21, 2026)
[^time-lord]: [Garrett Lord, TIME100 AI 2026](https://time.com/collection/time100-ai/2026/garrett-lord/) (TIME, 2026)
[^15]: [Some contractors who worked on OpenAI projects say Handshake AI is withholding thousands of dollars in pay](https://www.aol.com/articles/contractors-worked-openai-projects-handshake-050101409.html) (Business Insider via AOL, March 2026)
[^scale-dp]: [Scale AI data partnership](https://scale.com/data-partnership) (Scale AI, accessed October 2026)
[^tc-meta]: [Scale AI confirms significant investment from Meta, says CEO Alexandr Wang is leaving](https://techcrunch.com/2025/06/13/scale-ai-confirms-significant-investment-from-meta-says-ceo-alexandr-wang-is-leaving/) (TechCrunch, June 2025)
[^scale-ceo]: [Scale appoints new CEO](https://scale.com/blog/scale-appoints-new-ceo) (Scale AI, July 30, 2026)
[^tc-openai-scale]: [OpenAI drops Scale AI as a data provider following Meta deal](https://techcrunch.com/2025/06/18/openai-drops-scale-ai-as-a-data-provider-following-meta-deal) (TechCrunch, June 2025)
[^mercor-data]: [Mercor Data: monetize your data](https://www.mercor.com/data/) (Mercor, accessed October 2026)
[^tc-mercor-c]: [Mercor quintuples valuation to $10B with $350M Series C](https://techcrunch.com/2025/10/27/mercor-quintuples-valuation-to-10b-with-350m-series-c/) (TechCrunch, October 2025)
[^mercor-incident]: [Update on Mercor security incident](https://mercor.com/blog/update-on-mercor-security-incident/) (Mercor, June 25, 2026)
[^tc-mercor-month]: [After data breach, $10B-valued startup Mercor is having a month](https://techcrunch.com/2026/04/09/after-data-breach-10b-valued-startup-mercor-is-having-a-month) (TechCrunch, April 2026)
[^time-spirit]: [Google's $10 million bid for Spirit Airlines' data reveals AI's next frontier](https://time.com/article/2026/08/25/google-spirit-airlines-ai-data-RL/) (Time, August 2026)
[^lazarus]: [Project Lazarus](https://lazarus.turing.com/) (Turing, accessed October 2026)
[^tc-turing]: [Turing, a key coding provider for OpenAI and other LLM producers, raises $111M at a $2.2B valuation](https://techcrunch.com/2025/03/06/turing-a-key-coding-provider-for-openai-and-other-llm-producers-raises-111m-at-a-2-2b-valuation) (TechCrunch, March 2025)
[^sc-assethub]: [SimpleClosure AssetHub](https://simpleclosure.com/product/assethub/) (SimpleClosure, accessed October 2026)
[^sc-webull]: [SimpleClosure AssetHub on track to return $10M+ to venture-backed startups](https://www.webull.com/news/15515143679517696) (SimpleClosure release via Webull, 2026)
[^gizmodo]: [Failed companies are selling old Slack chats and email archives to train AI](https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916) (Gizmodo, citing Forbes, April 2026)
[^mit]: [AI labs are buying data from dead startups to train models](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models) (MIT Technology Review, October 13, 2026)
[^pulse2-protege]: [Protege: interview with co-founder and CEO Bobby Samuels about the AI data platform](https://pulse2.com/protege-profile-bobby-samuels-interview/) (Pulse 2.0, June 2026)
[^pulse2-a1]: [Protege: $30 million raised to expand real-world data access for AI development](https://pulse2.com/protege-30-million-raised-to-expand-real-world-data-access-for-ai-development/amp/) (Pulse 2.0, January 2026)
[^alleywatch]: [Protege and the market for AI training data licensing](https://alleywatch.com/2026/01/protege-ai-training-data-licensing-real-world-data-bobby-samuels/) (AlleyWatch, January 2026)
[^troveo]: [Troveo](https://www.troveo.ai/) (Troveo, accessed October 2026)
[^decoder-video]: [OpenAI and Google are buying YouTubers' unpublished videos for up to $4 per minute](https://the-decoder.com/openai-and-google-are-buying-youtubers-unpublished-videos-for-up-to-4-per-minute/) (The Decoder, citing Bloomberg, January 2025)
[^korra]: [Shutting down? Who buys your Slack, Jira and email data?](https://korra.io/guides/shutting-down-who-buys-slack-and-jira-data) (Korra, accessed October 2026)
[^ai26]: [AI26: turn your company's historical data into revenue](https://ai26.net/) (AI26, accessed October 2026)
[^datarade]: [Datarade provider plans](https://providers.datarade.ai/apply) (Datarade, accessed October 2026)
[^decoder-spree]: [Tech giants are on a billion-dollar shopping spree for AI training data](https://the-decoder.com/tech-giants-are-on-a-billion-dollar-shopping-spree-for-ai-training-data/) (The Decoder, citing Reuters, April 2024)
[^awsdx]: [AWS Data Exchange](https://aws.amazon.com/data-exchange/) (Amazon Web Services, accessed October 2026)
[^snowflake]: [Snowflake Marketplace documentation](https://docs.snowflake.com/en/user-guide/data-marketplace) (Snowflake, accessed October 2026)
[^databricks]: [Get started as a Marketplace provider](https://docs.databricks.com/aws/en/marketplace/get-started-provider) (Databricks documentation, accessed October 2026)
[^human-native]: [Cloudflare buys Human Native to strengthen licensed data access for AI developers](https://siliconangle.com/2026/01/15/cloudflare-buys-human-native-strengthen-licensed-data-access-ai-developers/) (SiliconANGLE, January 2026)
[^petrarch]: [Petrarch](https://www.ycombinator.com/companies/petrarch) (Y Combinator, accessed October 2026)
[^surge]: [Surge AI home page](https://surgehq.ai) (Surge AI, accessed October 2026)
[^appen-ots]: [Off-the-shelf datasets](https://datasets.appen.com/data-types-image-and-text) (Appen, accessed October 2026)
[^toloka-about]: [About Toloka](https://toloka.ai/about) (Toloka, accessed October 2026)
[^snorkel]: [Snorkel AI](https://snorkel.ai/) (Snorkel AI, accessed October 2026)
[^labelbox]: [Labelbox](https://labelbox.com/) (Labelbox, accessed October 2026)
[^invisible]: [Invisible AI training services](https://invisibletech.ai/ai-training) (Invisible, accessed October 2026)
[^reddit-reuters]: [Exclusive: Reddit in AI content licensing deal with Google](https://tech.yahoo.com/articles/exclusive-reddit-ai-content-licensing-041012503.html) (Reuters via Yahoo, February 2024)
[^ecfr]: [45 CFR 164.514, other requirements relating to uses and disclosures of protected health information](https://www.ecfr.gov/api/renderer/v1/content/enhanced/current/title-45?part=164&section=164.514) (eCFR, accessed October 2026)
[^gdpr26]: [Recital 26 GDPR, not applicable to anonymous data](https://gdpr-info.eu/recitals/no-26/) (gdpr-info.eu, unofficial reproduction, accessed October 2026)
[^20]: [How do we ensure anonymisation is effective?](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/data-sharing/anonymisation/how-do-we-ensure-anonymisation-is-effective/) (Information Commissioner's Office, accessed October 2026)
[^ftc2]: [AI (and other) companies: quietly changing your terms of service could be unfair or deceptive](https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2024/02/ai-other-companies-quietly-changing-your-terms-service-could-be-unfair-or-deceptive) (FTC Office of Technology, February 2024)

## FAQ

**What are the best platforms for selling data to AI companies?**

As of October 2026, the ten we rank are SellYourAIData, micro1, Handshake AI, Scale AI, Mercor, Turing (Project Lazarus), SimpleClosure Asset Hub, Protege, Troveo and Korra. Which is best depends on your situation: an operating company that wants competing offers, a company that is shutting down, a healthcare or media rights holder, or a team that wants a single quick payment. See the [comparison table](#all-10-platforms-compared).

**How much do AI companies pay for business data?**

Published ranges run from $10,000 at the low end of Scale AI's illustrative range to more than $4 million at the top of Handshake AI's. micro1 publishes tiers of $100,000, $500,000 and $1 million or more. Reported completed sales by shut-down startups were $10,000 to $100,000 per company, and Google's winning bid for Spirit Airlines' data was $10 million. Try our [free valuation estimator](/valuation/) for a range based on your own data.

**Is it legal to sell company data to AI companies?**

It can be, but it depends on what the data contains and what you promised the people in it. Customer contracts, privacy policies, employee notices, HIPAA, GDPR and state privacy laws can all limit a sale. Personal information usually has to be removed first. Read our [legal frameworks paper](/research/legal-frameworks-selling-business-data/) and have a lawyer review any agreement.

**What kind of data do AI companies buy?**

The categories buyers name most often are workplace messages and email, documents and procedures, tickets and project records, CRM records and source code. Healthcare records and video appear on specialized platforms such as Protege and Troveo. See our pages on [communications data](/data/communications/), [documents](/data/documents/), [CRM history](/data/crm-history/) and [code](/data/code/).

**Should I sell directly to a buyer program or use a broker?**

A buyer program is faster and sets its own price. A broker or seller-side representative costs a fee but can bring competing offers and negotiate terms such as exclusivity and field of use. The math is in our [direct vs broker guide](/guides/selling-direct-vs-broker/).

**Who removes personal information before the data is sold?**

It varies. Mercor, Scale AI, micro1, Turing and SimpleClosure say they de-identify the data as part of the process. Korra says it does not open, clean or deliver records, so that stays with you. Whoever does it, the contract should say who carries the risk if personal data slips through. See our [de-identification paper](/research/deidentification-enterprise-text/).

**Can I sell the same data to more than one AI company?**

Only if your first agreement is non-exclusive. A single payment with a full rights transfer, as some programs describe, may end your ability to license the same records again. Ask for non-exclusive terms or price exclusivity separately.

**Does SellYourAIData buy data?**

No. SellYourAIData represents the seller. We estimate the data's value, bring a summary to several qualified buyers and negotiate the terms. The owner prepares the data and remains responsible for removing personal information. We have no affiliation with any other company on this list, and our fee comes from a completed deal.