MarTech Verdict

Reviews · AI data licensing

Scale AI Data Partnership Review 2026: In-Environment Licensing

How Scale AI's data partnership program anonymizes inside your environment, which systems it takes, what it pays and its referral bonus, checked November 7, 2026

By the MarTech Verdict Editorial Team · Published November 7, 2026

How we put this together: We read Scale AI's data partnership page and FAQ, its security, about and terms pages, its June 12, 2025 announcement of Meta's investment and its July 30, 2026 announcement of a new CEO, all on November 7, 2026. Scale's application form: Couldn't confirm. We haven't sold data to Scale, applied for a partnership or spoken to the company for this article. Figures are Scale's own unless attributed.

Scale AI’s data partnership program turns a company’s de-identified operational records (chat, email, documents, tickets, CRM, HR and ERP data, and warehouse tables) into AI training data. Scale says it anonymizes everything inside your environment before extraction, approves every package with you and pays an illustrative $10K to $1M+ per partnership, per its page read November 7, 2026.

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

Scale AI is the San Francisco data-labeling company that supplies training data to AI labs. Its data partnership page is aimed at operators, and its pitch is different from most buyers here: “We are looking for partnerships, not transactions.” This review looks at what that means for the team that has to run it.

Verdict

Scale has one of the strongest published data-handling models of the buyers we’ve reviewed, for one reason: it says de-identification happens inside your environment, before anything is extracted, with per-package approval, an audit trail and on-premises or VPC options. Its security page lists recognized certifications, and it’s a large, established company.

What it doesn’t publish is the commercial side. There’s no payment timing, no split between upfront and recurring revenue, no exclusivity or retention terms and no named buyers. The value range is labelled illustrative. Treat Scale as a strong option for a company that wants to keep raw data inside its walls, and expect to negotiate the money terms from scratch.

Who it suits, and who should skip it

Scale may suit:

  • Companies with 40+ employees (its “preferred starting point,” assessed case by case) and years of records in mainstream SaaS tools.
  • IT and security teams that won’t let raw data leave, or need on-premises or VPC processing.
  • Owners who want to start with one workflow rather than a full corpus. Scale says partnerships “can begin with a very limited scope.”

Skip it, or go slowly, if:

  • You want a one-off sale with a known date of payment.
  • You need to know which AI companies will use your data.
  • You can’t name an internal owner. Scale wants “a clear owner who can scope the partnership, approve packages, and keep reviews moving.”

How Scale’s data partnership works

Scale frames the program as a continuous partnership: you contribute “data, records and workflows” and sign off on every data package, and Scale structures de-identified records into training data. It says it also advises partners on where AI can make their operations more efficient.

Its governance section lists what “never leaves”: employee and customer identities, raw systems, credentials or source access, and anything outside your signed agreement. Controls listed are de-identification, encryption in transit and at rest, per-package approval gates, a full audit trail and on-premises or VPC options. Scale says “0 bytes” are shared without your explicit, per-package approval.

On anonymization, the FAQ is specific: “Scale anonymizes everything inside your environment, before any data is extracted. PII is stripped and records are processed so they can’t be traced back to individuals or accounts.”

What Scale takes

Source typeSystems Scale namesWhat Scale says it keeps
Data warehousesBigQuery, Snowflake, DatabricksTable schemas and the structured records behind them
Chat and messagingSlack, Microsoft Teams, Google Chat“Full conversation histories, threads, participants, and reply structure”
Email and calendarGmail, Outlook, Google CalendarThreads, event details, senders, recipients, invitees and responses
Documents and filesWord, Docs, Slides, Excel, Sheets, Drive, SharePointFile contents plus metadata, authorship, timestamps and access permissions
Work and knowledgeJira, Confluence, Asana, Notion, LinearTickets, wikis and project records with complete history
Business platformsSalesforce, HubSpot, Workday, ServiceNow, SAP, ZendeskCRM, HR, ERP and support records

Scale lists about 30 industries of interest, from software, MSPs and marketing agencies to law firms, freight brokerage and 3PL, field services such as HVAC and plumbing, property management and staffing. Its FAQ says AI companies want to improve models “for certain corporate personas such as marketers or financiers,” not to compete with you.

Workday and HR records appear on the source list. Most buyers exclude HR data by default, so decide your position on HR, payroll and performance records before scoping.

Terms at a glance

Read November 7, 2026.

TermWhat Scale says
Stated value“$10K–$1M+ illustrative value per data partnership, scaling with cadence”
StructureContinuous partnership; “opportunity for recurring revenue”
Payment timingNot published
Eligibility40+ employees preferred; case by case
Starting scopeCan begin “with a very limited scope”; no full corpus required
De-identificationInside your environment, before extraction
ApprovalYour sign-off on every data package
SecuritySOC 2 Type II, ISO/IEC 27001:2022, DoD IL4 provisional authorization, FedRAMP High authorized (security page)
Buyers“AI companies”; none named
Exclusivity, retention, deletionNot published
Referral“Up to $50,000 bonus,” paid only if Scale signs a partnership agreement; one referral per deal

Who owns and runs Scale

Scale AI, Inc. is headquartered in San Francisco and was founded in 2016, per its about page, which also lists a $29B valuation, 1,000+ employees and “$1B Paid to contributors globally.”

On June 12, 2025, Scale announced “a significant new investment from Meta Platforms, Inc.” that valued it at over $29 billion, and said founder Alexandr Wang was joining Meta while remaining a director. Press reports at the time put Meta’s stake at about 49% and the investment at about $14.3 billion; Scale’s announcement gives neither figure. On July 30, 2026, Scale said it had appointed Francis deSouza as CEO, effective August 10, 2026, with Wang as founder and chairman.

For a data seller, the ownership matters in one way: a major AI developer is a large investor in the company that would hold and structure your data. Ask Scale which buyers your packages may reach and whether you can exclude any.

Risks

  • Undefined money terms. Value is illustrative, and payment timing and structure aren’t published. Get the payment trigger and date in the agreement.
  • Access inside your environment. In-environment anonymization means Scale’s tooling or staff operate inside your systems. Agree who gets access, for how long and how it’s logged.
  • Recurring scope. A continuing partnership means continuing data flows. Agree how to pause or end packages.
  • HR and client data. Workday and CRM records are on Scale’s list. Many 2026 agency contracts carry a “no client data used to train AI” warranty (RICHT law firm, October 18, 2026).
  • Buyer opacity. No buyers are named, and a major AI developer is a large investor.
  • Retention and trained models. Not published. Deleting data doesn’t remove it from models already trained on it.

Alternatives

  • Polyshares publishes a typical deal range of $100K to $2M and payment net 30 to 60 days, with an outside PII partner rather than in-environment processing.
  • micro1 publishes tiers from $100K to $3M+ and pays within 30 days of invoicing once it approves the data, but takes data before de-identifying it.
  • Endless Data says identifiers are stripped before anything leaves your company and agrees a fixed sum upfront.
  • Mercor is the connector-based contrast, paying typically 2 to 4 weeks after signing.
  • Handshake AI, Replay and Mode are reviewed separately.

See the comparison of companies that buy business data for AI for extraction models side by side, and the guide to selling company data to AI for the full project plan.

Before your first scoping call

  1. Pick one workflow and its systems, with years of history and record counts.
  2. Name the internal owner who will approve packages.
  3. Write your exclusions (HR, payroll, client workspaces, recorded calls) before Scale proposes a scope.
  4. Ask what software or staff run inside your environment, and on what access terms.
  5. Ask for the payment trigger, timing and how recurring revenue is calculated.
  6. Request exclusivity, retention, deletion and buyer terms in the draft agreement.
  7. Ask for the current SOC 2 report.

Deciding whether to sell at all? onfolio.com’s Scale AI data partnership review covers it from the owner’s side: what you might get, how a deal runs and what to ask before you sign.

Sources

Scale AI’s data partnership page

Think your company's a fit? Tell us about it and we'll introduce you.

Scale AI says its preferred starting point is 40+ employees, but don't let that stop you. Tell us about your company and, if it looks like a fit, we'll introduce you to Scale AI or whichever AI data buyer suits you best (others include Handshake AI, micro1, Mercor, Polyshares and Replay).

How this works: MarTech Verdict is owned by Onfolio. Onfolio reviews each submission and, where there's a fit, introduces you to a buyer, often through that buyer's referral program. Onfolio may receive a referral fee if a deal is signed. It doesn't charge you anything, and we only use these details to make introductions.

Main systems (tick all that apply)

Considering licensing your company’s data? Apply to Polyshares for an assessment.

FAQ

What is the Scale AI data partnership?
It's Scale AI's program for turning a company's de-identified operational workflows into AI training data. Scale says it anonymizes everything inside your environment before any data is extracted, approves each data package with you, and designs partnerships for recurring revenue. It shows an illustrative $10K to $1M+ per partnership and prefers companies with 40+ employees (read November 7, 2026).
Is Scale AI legit?
Scale AI, Inc. is an established San Francisco company founded in 2016. Its about page lists a $29B valuation and 1,000+ employees, and its security page lists SOC 2 Type II, ISO/IEC 27001:2022 and FedRAMP High authorization. Meta made a significant investment in June 2025, and Francis deSouza became CEO on August 10, 2026.
How much does Scale AI pay for company data?
Scale publishes only an illustrative range: "$10K–$1M+ illustrative value per data partnership, scaling with cadence." It doesn't publish payment timing or how upfront and recurring payments split. It says partnerships can start with a very limited scope, so a first package may sit at the low end.
Does Scale AI pay referral bonuses?
Yes. Scale's FAQ says you can "refer a company and earn up to $50,000 bonus," paid only if Scale signs a partnership agreement. Only one referral is paid per deal, to whoever successfully introduces Scale to the right stakeholder, and referrals must go through its official referral process.