DataPacts

Labelbox vs Scale AI

Checked

At a glance

Labelbox and Scale AI are two US training-data companies that both license company records for AI training, Labelbox through Prism and Scale through Scale Data Partnerships; they differ most in how a company is paid and what the license binds, since Labelbox states an upfront fee at signing plus a revenue share under licenses that are exclusive by default, while Scale states an illustrative value range and publishes no license terms.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

For a company weighing AI training data licensing, every term below is the company's own claim with a check date, here 2026-11-09. No seller payment under either program is on record in the sources checked.

No program wins by fees, and licensing to neither is a complete outcome. The page compares the two as buyers of company data and expert work, not their labeling software. Scale holds a contributor settlement and five GDPR complaints on record; Labelbox holds none in the sources checked.

Labelbox and Scale AI Side by Side, Dated

Each cell quotes the company's own page or a court or regulator record as checked on November 9, 2026, unless the row names a later re-check, with the source in the last column.

The table carries no score, rank or winner, and every value is the company's claim unless a court or regulator record is cited.

Attribute Labelbox (Labelbox Prism profile) Scale AI (Scale AI profile) Checked Source
Legal entity, country Labelbox, Inc. (Terms of Service; the Prism pages name no entity); United States Scale AI, Inc., a Delaware corporation; affiliate Smart Ecosystem, Inc. dba Outlier AI; United States 2026-11-16 / 2026-11-09 docs.labelbox.com terms of service; Outlier Terms of Use
Programs Prism (company data); Alignerr (contributors, ALIGNERR LLC) Scale Data Partnerships (company data); Outlier (contributors) 2026-11-09 program pages
Data wanted Reasoning traces, work-tool interaction records, support logs, medical records, robotics telemetry, code edit traces, consented likeness and voice, SAP data and financial workbooks Data warehouses, chat, email and calendar, documents, work tools, business platforms 2026-11-09 prism.labelbox.com and /faq; scale.com/data-partnership
Stated eligibility No explicit size rule; distressed and wound-down companies eligible; timestamped consent for media; IRB approval or qualified de-identification for medical data 40+ employees preferred, case by case; established systems, real workflows, domain depth, clear stakeholder owner 2026-11-09 program pages
Payment structure "an upfront licensing fee at signing plus a revenue share on downstream use"; split not published Not published; "opportunity for recurring revenue" 2026-11-09 prism.labelbox.com/faq; scale.com/data-partnership
Stated value Datasets "typically start around $50K"; large partnerships "can reach $1M+"; "$100M+" distributed and "1k+ companies licensing" (undated) "$10K–$1M+ illustrative value per data partnership, scaling with cadence" 2026-11-09 program pages
Stated timeline Scope about 10 minutes, processing 1 to 3 days, license 5 to 10 days; first wire "in under 21 days"; written quote within 48 hours and preliminary valuation within one business day, both on the home page Not published 2026-11-09 / 2026-11-16 prism.labelbox.com
Data preparation Not compared here Anonymization inside the company's environment before extraction; per-package approval; on-prem or VPC 2026-11-09 scale.com/data-partnership
License and deletion Exclusive by default; ownership kept; deletion within 30 days on kill-clause triggers; trained models "cannot be unlearned"; term, indemnity and audit not stated Exclusivity, term, deletion and retraining not published 2026-11-09 prism.labelbox.com/faq; scale.com/data-partnership
Receiving labs Not named Not named 2026-11-09 program pages
Referral terms None published Up to USD 50,000, only if Scale signs; one referral per deal 2026-11-09 / 2026-11-16 program pages
Funding or ownership statement Series E of more than USD 100 million incl. equity and a credit facility, USD 1.25 billion pre-money Meta minority stake, valuation over USD 29 billion 2026-10-14 / 2025-06-12 company blogs
Contributor platform terms Alignerr: "$80/hr" average and "$40-120/hr" listings claimed; independent contractors; no obligation to pay for the interview Outlier: weekly, hourly or per task, "Rates vary"; independent contractors; individual arbitration 2026-11-09 alignerr.com; outlier.ai
Court and complaint record None in the sources checked McKinney settled (final approval 2026-10-30); Schuster voluntarily dismissed; Ramey notice of settlement; 5 noyb GDPR complaints 2026-11-09 court and complaint table below

Where Prism and Scale Data Partnerships Differ for a Company Licensing Data

The two programs differ on four points a company feels before signing: how money arrives, what the license binds after signing, how fast and where the data is prepared, and which companies each program names as a fit.

An Upfront Fee Plus Revenue Share Against an Illustrative Value Range

Labelbox states that Prism pays "an upfront licensing fee at signing plus a revenue share on downstream use", with "Smaller, more targeted datasets" that "typically start around $50K", while Scale states "$10K–$1M+ illustrative value per data partnership, scaling with cadence", each checked 2026-11-09. Labelbox adds that large partnerships "can reach $1M+" and publishes no share percentage or share timing.

Its home page, read 2026-11-09, also claims "$100M+" distributed and "1k+ companies licensing", without a date.

Scale's range, USD 10,000 to USD 1,000,000 or more as stated on 2026-11-09, is an illustration, not a price list. Scale designs partnerships "so there's opportunity for recurring revenue" and publishes no payment timing or structure.

Neither program can be ranked on money: both ranges are company claims, not market prices, and no seller has confirmed a payment under either. Payment shapes in general are covered under upfront fee, revenue share and other ways AI data buyers pay.

Exclusive by Default With a Kill Clause Against Unpublished License Terms

Labelbox states that Prism licenses are exclusive by default while the seller keeps full ownership and continued use, whereas Scale publishes no exclusivity, term, deletion or retraining terms, as of 2026-11-09. Labelbox adds that buyers pay for stricter exclusivity.

Prism's FAQ describes a kill clause: deletion within 30 days after a material breach, a regulatory change or a change-of-control trigger the seller defines.

It also states that "Models already trained on the data cannot be unlearned" and that "no further training, transfer, or sublicensing is permitted post-revocation". The FAQ states no license term, indemnity or audit right.

Scale leaves duration, exclusivity, deletion and retraining to the signed agreement, and neither company names a receiving lab. The lock-in trade-off is set out under exclusive vs non-exclusive data licenses.

Stated Timelines Against Anonymization Inside the Company's Systems

Labelbox states a path from first call to first wire "in under 21 days", while Scale publishes no timeline and instead describes anonymizing data inside the company's own environment before any data is extracted. As each company describes it:

  • Prism's stages (home page, checked 2026-11-09): scoping about 10 minutes, processing 1 to 3 days, licensing 5 to 10 days.
  • Prism's valuation promises: the home page carries both a written quote within 48 hours and a preliminary valuation within one business day (re-read 2026-11-16).
  • Scale's preparation: anonymization in place, per-package approval, on-prem or VPC.
  • Scale's timing and costs: not stated.

The seller still decides what enters a package, a review covered under de-identification before licensing data.

Wound-Down Companies at Labelbox, 40 or More Employees at Scale

Labelbox states that Prism sets no explicit size rule and that distressed and wound-down companies are eligible, while Scale names 40 or more employees as its preferred starting point, decided case by case.

The Prism home page asks for media sets "with explicit, timestamped consent from real individuals" and IRB approval or qualified de-identification for medical data. Its categories run from reasoning traces to support logs, robotics telemetry and financial workbooks.

Scale asks for established systems, real workflows, domain depth and a clear stakeholder owner, naming warehouses such as Snowflake and platforms such as Salesforce and SAP.

Only Prism names closed companies as eligible, as of 2026-11-09; who signs for one is covered under selling data from a shut-down company.

Referral Terms for Introducers: None Published Against Up to USD 50,000

Labelbox publishes no Prism referral program, while Scale states "Refer a company and earn up to $50,000 bonus", paid only if Scale signs a partnership agreement with the referred company, each checked 2026-11-09.

Scale pays one referral per deal, reconfirmed on 2026-11-16, reviews referrer profiles "within 1–2 business days" and publishes no formula below the maximum.

For an adviser, the contrast is a stated ceiling against no offer, payable only on Scale's signature. Every published offer is listed under referral programs of AI data buyers.

Funding and Ownership: Labelbox's Series E and Meta's Minority Stake in Scale

Labelbox announced on 2026-10-14 a Series E of more than USD 100 million, including equity and a credit facility, at a USD 1.25 billion pre-money valuation, while Scale AI announced on 2025-06-12 that Meta would hold a minority of Scale's outstanding equity at a valuation of over USD 29 billion.

Labelbox's post names Infinitum Partners as lead and states that "Labelbox reached profitability this year".

Scale gives no percentage for Meta's holding, and Scale remains Scale AI, Inc.

A holder can ask either company which labs would receive the records, whether any end-user limit applies, and what a change of control means; Prism lets the seller define such a kill-clause trigger. Meta's own buying is covered under Meta as a data buyer.

Expert Contributors Behind Each Company: Pay Claims and Contract Terms

Both companies also acquire training data from paid individuals: Labelbox through Alignerr, an expert network "powered by Labelbox", and Scale through Outlier, which Scale operates with its affiliate Smart Ecosystem, Inc.

Alignerr contributors contract with ALIGNERR LLC; the corporate link to Labelbox, Inc. is not stated. The network is profiled under Alignerr contributor platform.

Alignerr claims "$80/hr" average pay and listings at USD 40 to 120 per hour, checked 2026-11-09. Its expert count differs by page: 100,000+ on alignerr.com, 2.6M+ on labelbox.com, and "more than 4 million people" in Labelbox's own network in the Series E post.

  • Alignerr terms: independent contractors; no obligation to pay for the interview; two conflicting FAQ answers on pay after a rejected deactivation appeal.
  • Outlier terms (effective 2026-05-11): independent contractors; work product assigned (s.4(b)); individual arbitration with a class waiver (s.10).

Outlier states that it pays weekly, hourly or per task, at rates that "vary by expertise, project complexity, and location". It claims USD 500M+ paid to experts on its home page and USD 600M+ paid out on its about page, both checked 2026-11-09. Scale's side is profiled under Outlier AI (Scale AI).

Court, Complaint and Incident Record of Labelbox and Scale AI

Scale's record holds a court-approved contributor settlement, a voluntarily dismissed contributor suit, a suit with a notice of settlement and five GDPR complaints filed in Europe, while no court case, breach notice or regulator action naming Labelbox, Prism or ALIGNERR LLC appears in the sources checked as of 2026-11-09. The table lists each event by date with its status as of 2026-11-09; allegations are not findings.

Date Company Event Forum / docket Status as of 2026-11-09 Source
n/a Labelbox None recorded: no court case, breach notice or regulator action naming Labelbox, Inc., Prism or ALIGNERR LLC n/a n/a Court and regulator sources checked
2025-01-17 Scale AI Schuster v. Scale AI, Inc. et al. filed (contributor class action, allegations) N.D. Cal. No. 3:25-cv-00620-JSC Voluntarily dismissed by notice dated 2026-08-10 (with prejudice as to named plaintiffs, without as to the putative class) Dkt. 34
2025-02-11 Scale AI Ramey v. Scale AI, Inc. et al.: joint notice of settlement (Dkt. 32); claims not recorded N.D. Cal. No. 3:24-cv-06999 HireArt stipulation for dismissal 2025-05-12 (Dkt. 36); no dismissal order on the docket feed docket feed
2026-10-27 Scale AI 5 GDPR complaints by noyb with Intérêt à Agir, DIPLab and Reversing.Works (Outlier contributor tracking, Persona biometric checks, access requests) Data protection authorities in France, Italy, the Netherlands and Austria No authority decision recorded noyb.eu
2026-10-30 Scale AI McKinney v. Scale AI, Inc. et al. (contributor misclassification class settlement), gross USD 12,500,000; payments reported on Form 1099, not W-2; class period 2020-12-10 to 2026-02-28 Superior Court of California, County of San Francisco, CGC-24-620481 Settled; final approval order filed 2026-10-30 class notice and final approval order

Every Scale record concerns contributors or their data, not a company data license. noyb's complaints concern how Outlier handles contributors' data in Europe. A settlement is not a finding of wrongdoing.

Labelbox's empty row reflects the court and regulator sources checked as of 2026-11-09, not a guarantee that no dispute exists. The full narratives sit under contributor lawsuits against human data platforms.

Which Program Fits, and When Neither Does

Neither program is better by fees; each one's own stated terms point to different companies, as of 2026-11-09, and some companies fit neither.

Prism's stated terms point to:

  • Money wanted at signing.
  • A distressed or wound-down company.
  • Content placed with one exclusive licensee.
  • Consented media or reasoning-heavy records.

Scale's stated terms point to:

  • 40 or more employees with established systems.
  • Raw data that must not leave before anonymization.
  • Records in warehouses, work tools and business platforms.
  • Openness to recurring deliveries.

Neither may suit:

  • A holder needing non-exclusive, published terms from day one.
  • A holder needing named receiving labs.
  • A holder needing deletion to reach trained models.

Programs beyond these two are weighed under choosing a buyer for your data.

Licensing to Labelbox or Scale AI: Benefits, Risks, Downsides and Alternatives

Licensing to either company turns records into income whose final size cannot be read in advance, because Labelbox publishes no revenue share percentage and Scale publishes no payment structure, timing or license terms.

Benefits

  • Prism: money at signing, ownership kept.
  • Scale: anonymization in place, approval per package.
  • Both publish dated program pages.

Risks

  • Prism: trained models cannot be unlearned; term, indemnity and audit unstated.
  • Scale: license terms unpublished.
  • Scale: contributor settlement, GDPR complaints and a Meta minority stake.

Downsides

  • Prism's default exclusivity blocks parallel licensing.
  • Prism's share percentage and Scale's payment timing are unpublished.
  • Neither names a receiving lab.

Alternatives

  • Another program, among them Handshake AI, micro1, Mercor, SuperAnnotate and Turing Project Lazarus.
  • An intermediary or sell-side adviser.
  • An independent introduction.
  • Internal use.
  • Not licensing at all.

Keeping the data is a complete outcome. Every option is set out under alternatives to selling data to AI labs.

Where This Comparison Sits Among the Buyer Pages

This page sets two buyers against each other; the profiles hold each company's full record, and the directory and tables cover every other program. The directory of who buys AI training data lists every buyer type.

How This Comparison Is Checked

Each cell is read from the company's own page, a court filing or a regulator record and dated; a cell changes only when the source has been read again.

Labelbox comes first only by URL order, the page carries no referral links, and both companies get the same method; see the DataPacts disclosure. The court summary is information, not legal advice.

Frequently Asked Questions

Which pays companies more, Labelbox or Scale AI?

Neither can be ranked: Labelbox states datasets from around USD 50,000 to USD 1 million or more and Scale an illustrative range per partnership, each checked 2026-11-09, and no seller has confirmed a payment under either.

Does either company pay upfront for company data?

Labelbox states an upfront licensing fee at signing plus a revenue share; Scale publishes no payment timing or structure, only that partnerships are designed with an opportunity for recurring revenue.

Can a company license the same records to Labelbox and Scale AI?

Prism licenses are exclusive by default, and buyers pay for stricter exclusivity; Scale publishes no exclusivity terms, so the signed agreements decide whether both are possible.

Does this page compare Labelbox and Scale AI labeling software?

No; it compares the two as buyers of company data and expert work.

Their labeling services appear only as rows among data labeling companies as buyers and partners of data holders.

Who else competes with Labelbox and Scale AI for company data?

Handshake AI, micro1, Mercor, SuperAnnotate and Turing run comparable programs.

Each side has its own page of options: Scale AI alternatives and Labelbox Prism alternatives.