
On 2026-11-09 Scale stated an illustrative value of USD 10K to USD 1M+ per data partnership and a referral bonus of up to USD 50,000, paid only if Scale signs; it publishes no license terms and no payment timing.
Meta holds a minority of Scale's outstanding equity, as Scale announced on 2025-06-12. Its court record holds three contributor lawsuits, one of which, McKinney v. Scale AI, reached final approval of a USD 12,500,000 gross settlement on 2026-10-30.
Every program figure here is Scale's own claim, checked 2026-11-09, never a confirmed earning. DataPacts is an independent reference on AI training data licensing that covers Scale.
Each claim below carries its source and check date, and how DataPacts sources and dates every figure is set out in the methodology.
What Scale AI Buys From Companies: Scale Data Partnerships
Scale AI licenses operating data that companies hold in warehouses, chat, email, documents, work tools and business platforms, and it anonymizes that data inside the company's own environment before extraction. Scale calls the program Scale Data Partnerships and describes it on scale.com/data-partnership, checked 2026-11-09. The data sources it names fall into six groups:
- Data warehouses: BigQuery, Snowflake and Databricks.
- Chat: Slack, Microsoft Teams and Google Chat.
- Email and calendar.
- Documents.
- Work tools: Jira, Confluence, Asana, Notion and Linear.
- Business platforms: Salesforce, HubSpot, Workday, ServiceNow, SAP and Zendesk.
The list describes records that already exist inside a business, such as tickets, threads and CRM entries, rather than new work commissioned from contributors. That separates the program from Scale's contributor business, which produces new data through paid individuals and is covered further down this page.
Eligibility is stated as a preference, not a rule. Scale names 40+ employees as a preferred starting point, decided case by case, and looks for established systems, real workflows, domain depth and a clear stakeholder owner, checked 2026-11-09. A smaller company is not excluded by the wording, but it sits outside the stated starting point.
Scale's program is therefore one of many answers to who buys AI training data from operating companies, aimed at firms with years of system history.
Scale AI Company Identity
Scale AI, Inc. is a Delaware corporation, as the Outlier Terms of Use state, and it contracts with Outlier contributors together with its affiliate Smart Ecosystem, Inc. The table lists each identity attribute with its source, as of 2026-11-09.
| Attribute | Value | Source |
|---|---|---|
| Legal name | Scale AI, Inc. | Outlier Terms of Use |
| Legal form | Delaware corporation | Outlier Terms of Use |
| State file number | not found in the sources checked | n/a |
| Website | scale.com | scale.com |
| Country | United States | Outlier Terms of Use |
| Type | human-data and data-labeling company | scale.com/data-partnership |
| Contributor-platform affiliate | Smart Ecosystem, Inc., doing business as Outlier AI; Terms of Use effective 2026-05-11 | Outlier Terms of Use |
| Ownership | Meta holds "a minority of Scale's outstanding equity" (announced 2025-06-12) | Scale AI blog |
| Leadership | Jason Droege named interim CEO (2025-06-12) | Scale AI blog |
| Founded and founders | not stated in the sources checked | n/a |
The Delaware file number, the founding year and the founders are not public in the sources checked, so none is given. The contracting party for a data partnership would be named in the agreement itself.
The Outlier Terms of Use, effective 2026-05-11, are the only primary document in the sources checked that states Scale's legal form, and they name both companies as contracting parties for contributors.
Scale Data Partnerships Terms, Dated
Scale publishes an illustrative value range, a referral bonus and an eligibility threshold for its data partnerships, but no license terms or payment timing; the table lists each stated term with its check date. All values are Scale's statements as of 2026-11-09.
| # | Term | Scale's stated value | Program | Checked | Source |
|---|---|---|---|---|---|
| 1 | Value per partnership | Scale states "$10K–$1M+ illustrative value per data partnership, scaling with cadence" | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 2 | Payment timing and structure | not published | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 3 | Recurring revenue | Scale states it designs partnerships "so there's opportunity for recurring revenue" | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 4 | License scope | exclusivity, term, deletion and retraining not published | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 5 | Anonymization location | Scale states it anonymizes inside the company's environment before any data is extracted | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 6 | Approval and deployment | Scale states per-package approval gates; on-prem or VPC options | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 7 | Eligibility | Scale states 40+ employees as a preferred starting point, case by case | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 8 | Data sources | Scale names warehouses, chat, email and calendar, documents, work tools and business platforms | Scale Data Partnerships | 2026-11-09 | scale.com/data-partnership |
| 9 | Referral bonus | Scale states "Refer a company and earn up to $50,000 bonus", only if Scale signs a partnership agreement | Scale Data Partnerships referral track | 2026-11-09 | scale.com/data-partnership |
| 10 | Referral limits and review | Scale states one referral paid per deal; referrer profiles reviewed "within 1–2 business days"; referrals can be submitted before approval | Scale Data Partnerships referral track | 2026-11-09 | scale.com/data-partnership/apply |
The rows show where Scale's page stops. It states what Scale wants, how Scale handles the data and what a referrer can earn, but nothing a seller would sign: no exclusivity rule, no license term, no deletion duty, no rule on retraining and no payment trigger.
Those clauses only appear once Scale sends an agreement, so the page cannot be compared clause by clause with a contract.
Read beside other buyers, the pattern is common rather than unique to Scale: most company data programs publish a range or a process and leave license clauses to negotiation.
The dated rows can be set against program terms compared across buyers, where each program's stated values carry the same check date, and where a missing value is shown as missing rather than estimated.
How a Scale Data Partnership Works: Scoping, In-Environment Anonymization, Package Approval
Scale describes a process that starts with an application or a scoping call and keeps the data inside the company's environment until each package is approved. The four stages, as Scale states them on 2026-11-09:
- Apply or book a scoping call. The company either submits an application or schedules a call to scope which systems and records are in play.
- Anonymize inside the company's environment. Scale states that it anonymizes the data in the company's own environment before any data is extracted.
- Approve each package. Scale describes approval gates per data package, so each batch is released separately.
- Choose a deployment option. Scale names on-prem and VPC options for where the work runs.
No stage covers payment. Scale publishes no point at which money moves, no invoice step and no delivery acceptance test, so the process ends, on the page, at approval and deployment.
Several duties stay with the seller whatever Scale's process does. The company still has to establish that it holds the rights to license the records, decide whether employees and customers whose messages appear in them must be told, and run its own review of what Scale's anonymization removes.
Scale's in-environment step is Scale's claim about its own tooling; the seller's own de-identification before licensing data still decides what the company can show a regulator or a customer later.
Scale AI Payouts and Referral Bonus: Claims and What Stays Unpublished
Scale states an illustrative value of USD 10K to USD 1M+ per data partnership, "scaling with cadence", checked 2026-11-09, and publishes no payment schedule. The wording on its page is "$10K–$1M+ illustrative value per data partnership", and the range is Scale's own claim, not a market price.
An illustrative range is not a quote. It spans two orders of magnitude, names no data type, volume or industry at either end, and does not say whether the top figure was ever reached.
A seller cannot read its own price from it; the value of company data depends on rights, uniqueness, volume, freshness and the buyer's use, which is why the value of company data is assessed on the holder's own records rather than on a program's banner range.
The phrase "scaling with cadence" and the statement that Scale designs partnerships "so there's opportunity for recurring revenue" point to repeated deliveries rather than a single sale, checked 2026-11-09. Neither says whether payment is upfront, per package, a license fee or a revenue share, or when it falls due. The table separates what Scale publishes from what it does not, as of 2026-11-09.
| Item | Published by Scale? | Where | Source |
|---|---|---|---|
| Value range | yes, as an illustrative value per partnership | program page | scale.com/data-partnership |
| Payment timing | no | n/a | scale.com/data-partnership |
| Payment structure (upfront or revenue share) | no; only "opportunity for recurring revenue" | program page | scale.com/data-partnership |
| Exclusivity | no | n/a | scale.com/data-partnership |
| License term | no | n/a | scale.com/data-partnership |
| Deletion | no | n/a | scale.com/data-partnership |
| Retraining | no | n/a | scale.com/data-partnership |
| Referral bonus | yes, up to a stated maximum | program page | scale.com/data-partnership |
| Headcount threshold | yes, as a preferred starting point | program page | scale.com/data-partnership |
The referral track has the most detailed published conditions. Scale states "Refer a company and earn up to $50,000 bonus" on 2026-11-09, payable only if Scale signs a partnership agreement with the referred company.
One referral is paid per deal, referrer profiles are reviewed "within 1–2 business days", and referrals can be submitted before the referrer is approved.
The bonus is a maximum, with no published formula for reaching it. Scale's track sits among other referral programs of AI data buyers, which differ mainly in the event that triggers the fee.
Meta's Minority Investment in Scale AI and What It Means for Data Holders
Scale AI announced on 2025-06-12 that Meta would hold a minority of Scale's outstanding equity, in an investment that valued Scale at over USD 29 billion.
The same announcement named two people: Alexandr Wang joins Meta and stays a director of Scale, and Jason Droege becomes interim CEO of Scale. Both are named here only as Scale's announcement names them.
The announcement leaves out what a data holder would most want to know. Scale states neither the stake percentage nor the investment amount, and figures that appear in press coverage are not Scale's statements, so this page repeats neither.
The sources checked for this profile also hold no Scale statement on how the investment changes the handling of customer or partner data, and no primary record of customers ending or pausing work with Scale afterwards.
Until a primary source says otherwise, the documented position is the one above: a minority stake, a valuation statement and a change of leadership.
Meta is also a buyer and developer of AI models in its own right. Its own data sourcing, crawler and publisher arrangements are separate from its stake in Scale, and Meta as a data buyer is profiled on its own page.
For a company considering a Scale Data Partnership, the investment turns into contract questions rather than conclusions. A seller can ask:
- Which entities and end users will receive the licensed data or models trained on it, named in the agreement?
- May the license restrict delivery to specific AI developers, or exclude some?
- Does Scale's confidentiality duty cover affiliates and shareholders, and does any shareholder have access rights to customer or partner data?
- What happens to the data and to derived models if ownership or control of Scale changes?
None of these is a published Scale term; Scale's page states no recipient list and no restriction on end users, checked 2026-11-09. They are points to settle in writing before signing.
Scale AI Reputation: Court Cases and GDPR Complaints, Dated
The public record on Scale AI holds three US lawsuits brought by contributors and five GDPR complaints filed in Europe, and only one, McKinney v. Scale AI, has ended in a court-approved settlement amount. The table lists each record by date, with its status as of 2026-11-09.
| Date | Record | Forum and docket | Status as of 2026-11-09 | Covered in depth on | Source |
|---|---|---|---|---|---|
| 2025-01-17 | Schuster v. Scale AI, Inc. et al. filed against Scale AI, Inc., Smart Ecosystem, Inc. and Outlier AI, Inc.; psychological-harm class action (allegations) | U.S. District Court, N.D. Cal., No. 3:25-cv-00620-JSC | voluntarily dismissed (see 2026-08-10) | contributor lawsuits page | complaint, Dkt. 1 |
| 2025-02-11 | Ramey v. Scale AI, Inc. et al. (Scale AI, Inc., Outlier AI, Inc., HireArt, Inc.): joint notice of settlement (Dkt. 32); class action, claims not recorded | U.S. District Court, N.D. Cal., No. 3:24-cv-06999 | settlement terms not public; HireArt stipulation for dismissal 2025-05-12 (Dkt. 36); no dismissal order on the docket feed | contributor lawsuits page | docket feed |
| 2026-08-10 | Schuster: notice of voluntary dismissal | U.S. District Court, N.D. Cal., No. 3:25-cv-00620-JSC | closed: with prejudice as to named plaintiffs, without prejudice as to the putative class; terminated 2026-08-20 | contributor lawsuits page | Dkt. 34 |
| 2026-10-27 | Five GDPR complaints by noyb with Intérêt à Agir, DIPLab and Reversing.Works over Outlier (trackers, Persona biometric checks, data minimization, unanswered access requests; allegations) | data protection authorities in France, Italy, the Netherlands and Austria | filed; no decision found | noyb complaints page | noyb |
| 2026-10-30 | McKinney v. Scale AI, Inc. et al. (Scale AI, Inc., Smart Ecosystem, Inc., HireArt, Inc. and named individuals): misclassification class settlement, class period 2020-12-10 to 2026-02-28 | Superior Court of California, County of San Francisco, Case No. CGC-24-620481 | settled: final approval filed 2026-10-30; gross USD 12,500,000; payments reported as non-wages on Form 1099 | McKinney settlement page | final approval order |
Allegations in these records are not findings, and none concerns a company data partnership.
Every record concerns contributors, the people paid through Scale and Outlier to produce human data, not companies that license records to Scale. McKinney alleged misclassification of contributors as independent contractors and settled with final approval on 2026-10-30. Schuster was voluntarily dismissed, not settled.
Ramey has a noticed settlement whose terms are not public. noyb states that Outlier has more than 900,000 freelance workers, and no authority decision was found as of 2026-11-09. Each record is registered with other data buyer incidents, dated and sourced, so Scale's record can be read beside those of other buyers.
Is Scale AI Legitimate as a Buyer of Company Data?
Scale AI is an identifiable company: Scale AI, Inc. is named as contracting party in published Terms of Use and as defendant in state and federal court records, and it publishes its company data program openly.
The documented basics are the entity and its Delaware form, the affiliate Smart Ecosystem, Inc., a program page with dated claims, and court dockets in San Francisco and the Northern District of California.
Other points stay open. Scale publishes no license terms, and the Delaware file number, the founders and any leadership change after the 2025-06-12 interim appointment are not in the sources checked.
The noyb complaints have no decision as of 2026-11-09. None of this makes Scale illegitimate; it means the agreement, not the program page, carries the terms a seller relies on.
A seller approached in Scale's name can check that the message comes from scale.com, that the agreement names Scale AI, Inc. as party, and that no payment or credentials are requested before signing. The wider checks for whether a data licensing email is legitimate apply to any outreach naming Scale.
Benefits, Risks, Downsides and Alternatives to Licensing Data to Scale AI
Licensing to Scale offers a published process with anonymization inside the company's own environment, but its license terms are not public and its value range is illustrative. The four lists draw only on the record as of 2026-11-09.
Benefits
- A published process: scoping, anonymization inside the company's environment, approval per package, on-prem or VPC options.
- A stated value range and recurring-revenue wording, as Scale's claims.
- An identifiable counterparty, Scale AI, Inc., with a known contracting structure.
Risks
- License scope, exclusivity, term, deletion and retraining are not published and must be negotiated.
- Meta holds a minority of Scale's equity, and Scale has published no statement in the sources checked on what that means for partner data.
- A contributor record that includes a settled misclassification class action and five GDPR complaints with no decision as of 2026-11-09.
Downsides
- The 40+ employee preference leaves smaller companies outside the stated starting point.
- No payment timing or structure is published, so cash flow cannot be planned from the page.
A Scale agreement would have to carry everything the page omits: named recipients, permitted uses, exclusivity, term, deletion on expiry, rules on retraining and the payment trigger. Each of these clauses of an AI data licensing agreement is where a seller's protection actually sits, and none of them is visible before Scale's draft arrives.
Alternatives
- Other company data programs, among them Handshake AI Data Partnerships, Mercor Data Monetization, micro1 Company Data Partnerships and Turing Project Lazarus, in alphabetical order.
- A direct license with an AI lab.
- Keeping the data for internal use, including internal model work.
- Not licensing at all.
Scale AI Next to Other Company Data Programs
Scale AI is one of several human-data companies that also run a program for licensing company data. Scale Data Partnerships is one of the company data partnership programs run by such companies, each with its own stated range, thresholds and referral terms, checked 2026-11-09. The comparisons below set Scale against named programs:
- Mercor vs Scale AI on company data terms.
- micro1 vs Scale AI side by side.
- Scale AI vs Surge AI on programs and reputation.
- Scale AI alternatives for licensing company data.
Licensing to any program is one option among several, and the full set of alternatives to selling data, including keeping it, is weighed separately.
More About Scale AI
Scale AI Jobs, Outlier and Remotasks
Scale AI's contributor work runs through Outlier, which Scale operates through Smart Ecosystem, Inc., and pays weekly on Tuesdays, as Outlier states (checked 2026-11-09). Contributor jobs and pay across platforms are compared under AI training jobs and platforms. The profile of Outlier AI covers its pay, terms and searches for Remotasks.
Where Scale AI's Court and Complaint Records Are Tracked
Each record in the table above has a fuller page, and this summary is information, not legal advice.
- McKinney v. Scale AI settlement: settled, final approval filed 2026-10-30.
- Schuster and Ramey among contributor lawsuits: dismissed 2026-08-10 and settlement noticed 2025-02-11.
- noyb complaints against Scale AI: filed 2026-10-27, no decision as of 2026-11-09.
- Outlier's ID checks and tracking under contributor privacy, as of 2026-11-09.
Scale AI's Valuation Statement, Interim CEO and Founders
Scale's only valuation statement in the sources checked is "over $29 billion", given with the Meta investment on 2025-06-12. It is Scale's own figure, tied to that investment, and not a market price or a later valuation.
The same announcement named Jason Droege interim CEO; no later leadership change is recorded in the sources checked, so the interim appointment is the latest position stated here.
The founding year and the founders are not stated in the sources checked and are not given. Scale's revenue, profitability and listing status are not public in those sources either; funding rounds and valuations across the sector are covered under funding and valuations of AI data companies.
Scale AI Beyond Company Data: Human Data and RL Environments
Scale's role as a human-data supplier predates its company data program: the InstructGPT paper of March 2022 describes about 40 contractors hired "on Upwork and through ScaleAI". That work, paying people to write and rank model outputs, is the business Outlier now carries.
Scale is one of the companies that supply human data to AI developers. It also sells RL environments: Scale states 2,000+ tasks and four enterprise environments, checked 2026-11-09, and is listed with other RL environment companies.
Frequently Asked Questions About Scale AI
What does Scale AI do?
Scale AI supplies human-made training and evaluation data to AI developers and licenses anonymized company data through Scale Data Partnerships. Its contributor work runs through the Outlier platform, operated with its affiliate Smart Ecosystem, Inc.
Is Scale AI a subsidiary of Meta?
Scale states that Meta holds a minority of its outstanding equity, announced 2025-06-12. A minority holding does not make Scale a subsidiary. Scale states no percentage, and none is given here.
Who is the CEO of Scale AI?
Scale named Jason Droege interim CEO on 2025-06-12; no later change is on record here. The same announcement said Alexandr Wang joins Meta and stays a Scale director.
How much does Scale AI pay for company data?
Scale publishes only an illustrative value per data partnership, checked 2026-11-09, and no price, payment timing or payment structure. The range and its wording are set out in the payouts section above; it is not a quote.
Does Scale AI see a company's raw data?
Scale states that it anonymizes data inside the company's environment before any extraction, with approval per package. That is Scale's description of its process, checked 2026-11-09, not an independent audit.