Which companies buy data for AI training?

As of October 21, 2026, companies publicly offering to buy or license business data for AI included micro1, Scale AI, Mercor, Handshake AI, Labelbox Prism, Polyshares, Endless Data, Troveo and SimpleClosure AssetHub. Some are AI training-data companies that build datasets for labs, some resell or license on your behalf, and SimpleClosure focuses on startups, including those winding down.

How much do AI companies pay for data?

Advertised ranges on October 21, 2026 ran from $10K (Scale AI's illustrative low end) to "$100K to $4M+" (Handshake AI). Checkable deals are smaller: about $25,000 for one closed startup's code base (MIT Technology Review), $110,000 to $450,000 in Polyshares' unnamed case studies, and Google's $10M agreement for bankrupt Spirit Airlines' archive.

Can I get paid for training AI?

Companies can, by licensing operational records such as SOPs, tickets, CRM history and internal chat to the buyers compared here. Individuals have separate routes: several of these companies hire contractors, and micro1, Scale AI and Handshake AI pay referral bonuses of $50,000 or more, up to $50,000 and $10,000 respectively for introducing a company that signs.

What are some companies that sell data?

In the AI training market, Endless Data and Polyshares license company data and supply it on to AI labs and training-data companies, and Troveo licenses data on owners' behalf to more than 40 buyers, it says. SimpleClosure AssetHub sells or licenses startup code and workspace data to a vetted buyer network.

Companies That Buy Business Data for AI in 2026: 9 Compared

Nine buyers of company data for AI compared on systems taken, extraction, de-identification, retention, exclusivity and payment, checked October 21, 2026

At least nine companies publicly offered to buy or license business data for AI on October 21, 2026: micro1, Scale AI, Mercor, Handshake AI, Labelbox Prism, Polyshares, Endless Data, Troveo and SimpleClosure AssetHub. For ops teams, the biggest difference is extraction: Mercor connects to your tools, Scale anonymizes inside your environment, and Troveo works from your exports.

Onfolio may earn a referral fee if you sign up to Polyshares through our links. This does not affect our editorial assessment.

This comparison is about data handling: which systems each buyer takes, how data leaves your environment, who de-identifies it and where, and what happens to it afterwards. Pay and timing are in a separate, shorter table. Detailed reviews of five of these programs are linked in each section below.

Data handling compared

Everything below is each company’s own published claim, read October 21, 2026 (micro1 re-read October 25, 2026). “Not published” means we couldn’t find it on the company’s pages that day.

Buyer	Type	Systems or data named	Extraction method	Who de-identifies, and where	Sample review	Retention and deletion	Security claims
Endless Data	Pools data and licenses it to several buyers	Industry workflows; systems fixed in a manifest	Not published; you approve a manifest	Identifiers stripped before anything leaves your company	Representative samples	Deletion at end of term; its buyer terms let buyers keep trained models	Audit rights
Handshake AI	Connects businesses with AI labs	Not published by system	Not published (“prepare the data” step)	De-identification promised; who and where not published	Team approval before committing; review of who may receive data	Not published	Not published
Labelbox Prism	Licensing, upfront fee plus revenue share	Operational data; HR, reviews, compensation, benefits excluded by default	Data leaves your environment after PII removal	Prism removes PII first; names, emails, internal IDs tokenized before egress	Not published	Kill clause: deletion within 30 days on defined triggers; trained models “cannot be unlearned”	Encryption with customer-managed keys; regional residency
Mercor	Licensing platform	Slack, Teams, Gmail, Confluence, Salesforce, Outlook, GitHub, Drive, Notion; “100+ supported tools”	Connect your tools; platform extracts	Fully automated pipeline; no lab buyer sees raw data	Not published	Not published	SOC 2 Type II
micro1	Data partnership for RL environments	Internal communications, SOPs, project histories, CRM records, support tickets	Transfer through micro1’s APIs or other approved methods	No anonymization needed before transfer; micro1 de-identifies, PII removed “when relevant”; some synthetic rewrites	Representative samples before use	Raw data deleted on an agreed schedule; deletion on request or at end, per terms	Data segregation; limited authorized access; U.S. and allied buyers only
Polyshares	Buys licenses directly, supplies labs	Slack, Teams, email, Drive, SharePoint, spreadsheets, CRM, tickets, design files, call recordings, accounting and ERP	Not published; nothing leaves before an accepted offer	Unnamed specialist PII partner, before Polyshares or any lab sees it	Polyshares reads an anonymized sample to price	“Defined term” on one page; case studies “perpetual”	Two named staff see identity with data
Scale AI	Continuous partnership	Warehouses, Slack, Teams, Google Chat, Gmail, Outlook, Calendar, Docs, Drive, SharePoint, Jira, Confluence, Asana, Notion, Linear, Salesforce, HubSpot, Workday, ServiceNow, SAP, Zendesk	Per-package approval; on-prem and VPC options	Scale anonymizes inside your environment, before extraction	Per-package approval	Not published	Encryption; audit trail
SimpleClosure AssetHub	Facilitates sales and licenses of startup assets	Source code; Slack, Google Drive, email, Notion, Figma	Short form or GitHub connection	Every dataset processed for PII before licensing; where not published	Not published	Not published	Not published
Troveo	Licenses on your behalf, revenue share	Chat, email, docs, project tools, CRM, IT, HR and ops tickets, finance, warehouses, code	Your team’s exports; no OAuth or integrations	Troveo pseudonymizes to “one documented standard”	Not published	Not published	Not published

Three extraction models show up. Connector-based (Mercor) means a platform pulls from your tools with your authorization. In-environment (Scale) means anonymization runs before data leaves. Transfer-based models (Troveo from your team’s exports, Prism after PII removal, micro1 through its APIs or other approved methods) leave the question of exactly when raw data leaves your control, so ask.

Commercial terms in brief

Read October 21, 2026 (micro1 re-read October 25, 2026). This table is deliberately short; the price evidence page covers value in detail.

Buyer	Stated pay	Headcount floor	Payment timing	Exclusivity
Endless Data	“Up to $4M”	Roughly 30+	Fixed sum agreed before sharing; timing after signing not published	Non-exclusive
Handshake AI	“$100K to $4M+”	20+ full-time, 3+ years (referral page)	Not published	Not published
Labelbox Prism	About $50K to $1M+	None published	Upfront fee at signing plus revenue share	Exclusive by default
Mercor	No range published	None published	Wire on delivery to buyer, typically 2 to 4 weeks after signing	Not published
micro1	$100K-500K, $500K-2M, $2M+ tiers	30+ full-time	Up to 30 days to approve, then paid within 30 days of invoice	Not published
Polyshares	$100K to $2M+	None (“any company with real operating history”)	Net 30 to 60 after approval and delivery	Exclusive by default; case studies perpetual
Scale AI	$10K to $1M+ “illustrative”	40+ preferred	Not published	Not published
SimpleClosure AssetHub	“Tens to hundreds of thousands”	Startups, operating or closing	Not published	Not published
Troveo	Full-company deals “start at six figures”	None published	Revenue share per sale; timing not published	Time-limited, “earned”

The buyers, one by one

Endless Data

Endless Data is the trading name of Seldon Ventures, L.L.C., a Delaware LLC. It pools many companies’ data and licenses it non-exclusively to buyers it names as micro1, Mercor and Handshake, which are training-data companies rather than frontier labs. Its published Terms and Conditions are written for buyers, and we couldn’t find seller-side terms. Endless Data review.

Handshake AI

Handshake AI says it connects businesses with AI labs and advertises “$100K to $4M+.” Its process runs from a first conversation through agreeing scope and terms, preparing the data, and delivery and payment, with team approval before committing. It doesn’t publish extraction, retention or payment timing.

Labelbox Prism

Prism says smaller datasets start around $50K and large partnerships reach $1M+, paid as an upfront fee at signing plus a revenue share. It removes PII before data leaves your environment, encrypts with customer-managed keys and excludes HR and compensation data by default. Every license has a kill clause forcing deletion within 30 days on defined triggers, though Prism says models already trained “cannot be unlearned.”

Mercor

Mercor connects to your tools (it names Slack, Teams, Gmail, Salesforce, GitHub, Drive and Notion among “100+”) and runs a fully automated extraction and anonymization pipeline, stating SOC 2 Type II. It pays by wire on delivery to the buyer, typically 2 to 4 weeks after signing, and says connecting more tools generally raises the payout. Business Insider (October 13, 2026) reported that former Mercor gig workers allege in an ongoing lawsuit that Mercor encouraged them to upload work materials without employer consent; Mercor has denied the claims in court.

micro1

micro1’s Enterprise Data Partnership wants internal communications, SOPs, project histories, CRM records and support tickets from companies with 30+ full-time employees, and The Next Web (October 9, 2026) reported a $1bn commitment to buy company data over 12 months. Data arrives through its APIs or other approved methods without prior anonymization; micro1 removes PII “when relevant,” keeps data for agreed periods and pays within 30 days of invoicing once it approves the data. It doesn’t publish exclusivity. micro1 data partnership review.

Polyshares

Polyshares says it buys licenses directly and supplies AI labs, with an unnamed specialist PII partner anonymizing data before Polyshares sees it. Licenses are exclusive by default, and its four published case studies are perpetual exclusive licenses from $110,000 to $450,000. It doesn’t disclose which labs take which data. Polyshares review.

Scale AI

Scale pitches a continuous partnership and says it anonymizes everything inside your environment before extraction, with per-package approval, an audit trail and on-prem or VPC options. Employee and customer identities, raw systems and credentials never leave, it says. It prefers companies with 40+ employees and publishes an illustrative $10K to $1M+ per partnership.

SimpleClosure AssetHub

SimpleClosure, a startup wind-down service, sells or licenses source code and workspace data from Slack, Drive, email, Notion and Figma, for closing and operating companies. It excludes sensitive employee, compensation and health information and takes an unpublished percentage only on success. MIT Technology Review (October 13, 2026) reported over 150 deals. SimpleClosure AssetHub review.

Troveo

Troveo, which began in video licensing, licenses company data on the seller’s behalf and shares revenue on every sale. It works from your team’s exports with no OAuth or integrations, pseudonymizes to a documented standard, and offers time-limited exclusivity that “has to be earned.” Troveo review.

What data you can and can’t license

Across these buyers, the published exclusions agree on the hard lines: client data, personal data and HR records. The split for an ops team looks like this.

Usually firm-owned and licensable once scrubbed	Usually not yours to license, or high risk
Your SOPs, playbooks, templates and internal knowledge base	Client deliverables, client ad accounts and client ESP lists
Internal tickets and project history, client names removed	Client Slack Connect channels and shared client drives
Your own campaign process and CRM pipeline records, contacts stripped	Contact-level CRM personal data
Internal chat about how work gets done, de-identified	HR systems, performance reviews, compensation, health data
Design files and assets you made for yourself	Call recordings, live credentials, third-party licensed data

The rules behind the right-hand column:

Client data. Polyshares’ license doesn’t cover what you hold for clients; Endless never licenses client personal data, privileged matter content or client books. A service provider’s access to client data doesn’t authorize monetizing it, and many 2026 agency contracts carry a “no client data used to train AI” warranty (RICHT law firm, October 18, 2026). Mercor’s industry examples include client decks, SOWs and case files, which need the client’s rights.
Personal data. CCPA/CPRA has covered employee and B2B contact data since January 1, 2023. GDPR covers EU and UK people.
Employee chat. NLRA Section 7 protects employee discussions of pay and conditions.
Recorded calls. Consent rules differ by state, and Brewer v. Otter.ai (August 2025) alleges wiretap violations.
Platform terms. Slack’s API terms (August 2025) bar bulk export and LLM training via the API, per Hunton. Google Workspace API policies restrict training non-personalized models. This matters most for connector-based extraction.
Your privacy notice. The FTC said in February 2024 that retroactively changing privacy terms to allow AI training can be deceptive.

Risks for the ops team

Raw data in transit. Scale says it anonymizes inside your environment, and Endless says identifiers are stripped before anything leaves your company. micro1 says you don’t need to anonymize before transfer, and other buyers either don’t say where scrubbing happens or use a separate partner, so raw data may move before de-identification.
Exclusivity. Polyshares and Prism are exclusive by default. Epoch AI’s interviews (January 12, 2026) found exclusive deals for AI training environments priced 4 to 5 times non-exclusive ones, so ask what exclusivity adds.
Trained models persist. Prism says trained models can’t be unlearned, and Endless’s buyer terms let buyers keep trained models after deletion.
Buyer opacity. Polyshares doesn’t disclose which labs take which data, and MIT Technology Review says founders have little visibility into buyers.
Payment after work. Most buyers pay after delivery and acceptance, so your team does the export first.
Pressure tactics. Business Insider (October 13, 2026) reported 48-hour “exploding offers” are common in this market, without naming a company here. SimpleClosure (October 1, 2026) warns about new buyers advertising $1M to $5M.
Stolen or unlicensed data. Surge AI’s Nick Heiner told Business Insider a group tried to sell a codebase it had no rights to, which is why buyers check rights.
Diligence and reputation. A prior license and its exclusivity or change-of-control terms surface in M&A diligence (Frankfurt Kurnit via Mondaq, October 2026). Spirit Airlines’ sale drew a union objection and a letter from 100+ members of Congress (Business Insider).

Selection checklist for ops

Use these questions to compare buyers on the same terms. For the whole project, from data audit to delivery and payment, see our guide to selling company data to AI.

Ask whether extraction uses connectors, your exports or in-environment processing.
Confirm who de-identifies the data, with what method, and whether it happens before data leaves your systems.
Request a de-identified sample of your own data before signing.
Get the list of systems, years and record types in scope written into the agreement.
Ask for named end buyers, or at least notice of each buyer.
Request the retention period, a deletion certificate and the position on models already trained.
Ask for a SOC 2 report and a sub-processor list.
Compare exclusivity terms and ask what each buyer pays for exclusivity.
Confirm the payment date relative to signing, delivery and acceptance.

Deciding whether to sell at all? onfolio.com covers the same buyers from the owner’s side in companies that buy business data for AI: what you might get, how payouts work and what can go wrong.

Sources

Endless Data homepage and Terms and Conditions, read October 21, 2026: https://endless-data.com
Handshake AI data partnerships page, read October 21, 2026: https://joinhandshake.com/ai/data-partnerships
Handshake AI referral page, read October 21, 2026: https://joinhandshake.com/ai/data-partnerships/referral
Labelbox Prism FAQ, read October 21, 2026: https://prism.labelbox.com/faq
Mercor data page, read October 21, 2026: https://www.mercor.com/data/
micro1 data partnership page, read October 21 and 25, 2026: https://www.micro1.ai/data-partnerships
Polyshares homepage and data-handling page, read October 21, 2026: https://polyshares.com/data-handling
Scale AI data partnership page, read October 21, 2026: https://scale.com/data-partnership
SimpleClosure AssetHub page, read October 21, 2026: https://simpleclosure.com/product/assethub/
SimpleClosure, “How Real Are Those Million-Dollar Data Offers?”, October 1, 2026: https://simpleclosure.com/blog/guides/how-real-are-those-million-dollar-data-offers/
Troveo business data page, read October 21, 2026: https://www.troveo.ai/business-data
Business Insider, “AI labs are fueling a wild market for your Slack DMs,” October 13, 2026: https://www.businessinsider.com/ai-training-employee-data-emails-slack-dms-2026-10
MIT Technology Review, “AI labs are buying data from dead startups to train their models,” October 13, 2026: https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/
The Next Web, “micro1 commits $1bn to buy company data for training AI agents,” October 9, 2026: https://thenextweb.com/news/micro1-1bn-enterprise-data-ai-training-citi-hercules
Epoch AI, “An FAQ on reinforcement learning environments,” January 12, 2026 (interview-based): https://epoch.ai/gradient-updates/state-of-rl-envs
RICHT law firm on agency client data, October 18, 2026
Hunton on Slack API terms, August 2025
Frankfurt Kurnit via Mondaq on data licenses in M&A, October 2026
FTC guidance on retroactive privacy policy changes, February 2024
Brewer v. Otter.ai complaint, August 2025

FAQ

Which companies buy data for AI training?: As of October 21, 2026, companies publicly offering to buy or license business data for AI included micro1, Scale AI, Mercor, Handshake AI, Labelbox Prism, Polyshares, Endless Data, Troveo and SimpleClosure AssetHub. Some are AI training-data companies that build datasets for labs, some resell or license on your behalf, and SimpleClosure focuses on startups, including those winding down.
How much do AI companies pay for data?: Advertised ranges on October 21, 2026 ran from $10K (Scale AI's illustrative low end) to "$100K to $4M+" (Handshake AI). Checkable deals are smaller: about $25,000 for one closed startup's code base (MIT Technology Review), $110,000 to $450,000 in Polyshares' unnamed case studies, and Google's $10M agreement for bankrupt Spirit Airlines' archive.
Can I get paid for training AI?: Companies can, by licensing operational records such as SOPs, tickets, CRM history and internal chat to the buyers compared here. Individuals have separate routes: several of these companies hire contractors, and micro1, Scale AI and Handshake AI pay referral bonuses of $50,000 or more, up to $50,000 and $10,000 respectively for introducing a company that signs.
What are some companies that sell data?: In the AI training market, Endless Data and Polyshares license company data and supply it on to AI labs and training-data companies, and Troveo licenses data on owners' behalf to more than 40 buyers, it says. SimpleClosure AssetHub sells or licenses startup code and workspace data to a vetted buyer network.