Scale AI data partnerships: what it buys, who qualifies, and what it pays

What is Scale AI? Company background

Scale AI is a San Francisco company, founded in 2016, that supplies training data and model evaluation to AI developers (source: Scale AI, about, Fortune, May 21, 2024). It describes itself as an AI applications and data infrastructure company whose data engine supports model development at many frontier labs, and it also sells a generative AI platform to governments and large companies (source: Scale AI blog, new CEO).

Wang, a 19-year-old who had left MIT, and Guo first planned a concierge service for doctors. After joining Y Combinator in the summer of 2016 they pivoted to data labeling, and found their first big customers among self-driving car startups (source: Fortune, Jun 22, 2025). Guo left in 2018 and Wang became CEO (source: Fortune, Jun 22, 2025). Scale moved on to language models, helping OpenAI train GPT-2 in 2019, and took on defense and government work (source: Fortune, May 21, 2024).

In June 2025 Meta agreed to pay about $14.3 billion for 49% of Scale, a deal that valued it at over $29 billion. Wang left to join Meta's superintelligence effort and stayed on Scale's board, and chief strategy officer Jason Droege became interim CEO (source: TechCrunch, Jun 13, 2025, Fortune, Jun 13, 2025). Fortune reported that Meta would buy at least $500 million a year of Scale data for five years (source: Fortune, Jun 22, 2025). On July 30, 2026, Scale named Francis deSouza, formerly COO of Google Cloud and CEO of Illumina, as CEO effective August 10 (source: Scale AI blog, new CEO). The company-data program is a separate line from the contractor work most Scale reviews describe. It lives on its own page, scale.com/data-partnership, and Scale publishes no launch date or program lead for it (source: Scale AI, data partnerships).

Selling company data to Scale AI: how the program works

Scale frames the program as “partnerships, not transactions”: an ongoing arrangement scoped to how your business works, rather than a single archive handed over once (source: Scale AI, data partnerships). Based on its page, the steps are:

1. Apply. You start an application or book a scoping call. Scale invites you to “bring one workflow” to begin.
2. Scope. Scale shapes a continuous partnership around your business. Its FAQ says you do not have to start with your whole workspace and that partnerships can begin with a very limited scope.
3. Package and approve. Records are de-identified inside your environment and grouped into packages. Nothing moves without your sign-off on each package.
4. Get paid, and advised. Scale says the partnership is designed to pay back over time, and that it will advise you on where AI can make your own operations more efficient.

Scale's FAQ also answers a question many owners ask first. It says the AI companies are not using your data to compete with you, but to make their models more useful for roles inside your company, such as a marketer or someone on your finance team (source: Scale AI, data partnerships). Scale does not name which AI companies receive the data.

What Scale AI buys

Scale wants “structured, de-identified records” from the everyday systems your teams run on, and says it never takes the raw systems themselves (source: Scale AI, data partnerships):

It names five industry groups of interest, with 22 sub-industries between them (source: Scale AI, data partnerships):

That is the broadest industry list of the four large buyer programs we profile. Field services, equipment rental and hospitality do not appear on micro1's, Mercor's or Handshake AI's lists. If yours is not here, the directory of companies that buy data for AI shows which programs name your industry.

Who qualifies for a Scale AI data partnership

Scale's size bar is the highest of the four large buyer programs we profile. Handshake AI starts at 20+ full-time employees and 3+ years, micro1 at 30+, and Mercor publishes no minimum (source: Handshake AI, license your data, micro1, data partnerships, Mercor, Monetize your data). The internal-owner point is worth taking seriously: a partnership with package-by-package approval needs someone on your side with the time and authority to say yes or no.

What Scale AI advertises paying

Scale publishes a single figure: $10K to $1M+, labeled an “illustrative value per data partnership, scaling with cadence” (source: Scale AI, data partnerships). Its FAQ adds that partnerships are designed so there is opportunity for recurring revenue. It does not define cadence, publish tiers, say what moves a partnership toward the top of the range, or publish a payment schedule or any fees.

Two things follow. First, the low end is well below what other large buyers advertise as a starting point: micro1's first tier is $100K to $500K and Handshake AI says its payments to date have ranged from $100K to $4M (source: micro1, data partnerships, Handshake AI, data partnerships). Scale's range may reflect that it starts small, with one workflow, and grows. Second, a recurring structure can be worth more or less than a single payment depending on how long it runs, how often you deliver, and whether Scale can end it. You can only judge it against a concrete schedule.

Those are advertised figures, not offers. Our company data value checker shows which programs you likely qualify for and what each one advertises.

How Scale AI says your data is handled

Scale makes these statements about governance on its data partnership page (source: Scale AI, data partnerships):

• Scale anonymizes everything inside your environment, before any data is extracted.
• Personal information is stripped and records are processed so they cannot be traced to individuals or accounts.
• You sign off on every data package, and nothing is shared without that approval.
• Data is encrypted in transit and at rest, with a full audit trail.
• On-prem and VPC options are available.
• Employee and customer identities, raw systems, credentials and source access, and anything outside the signed agreement never leave.

De-identifying inside your environment is a real difference from programs that extract first and clean later. It also means Scale's tooling runs on your systems, so ask what access it needs, who on its side can see what, and how that access ends. Scale's page does not mention exclusivity, deletion of delivered data at the end of a partnership, or a security certification.

Scale AI data partnerships review: what stands out

Measured against its own page and the other large buyer programs, here is how the terms read.

• Strongest on governance. In-environment anonymization, per-package approval and on-prem options are spelled out in detail (source: Scale AI, data partnerships).
• Broadest industry list. Five groups and 22 sub-industries, including field services and property management (source: Scale AI, data partnerships).
• Lightest on commercial terms. No payment schedule, fees, exclusivity or written referral terms, and two different referral amounts on one page (source: Scale AI, data partnerships, Scale AI, data partnership referral form (page script)).
• A long track record as a company. Founded in 2016, with a 2025 valuation of $29 billion (source: Wikipedia, Scale AI, TechCrunch, Jun 13, 2025). The data partnership program itself has no published launch date.

What to get in writing before you sign

Because Scale publishes so few commercial terms, the written agreement carries more weight than usual. Ask for each of these:

1. Price per delivery and the cadence. How often you deliver, what each package pays, and the minimum you are guaranteed if Scale stops taking deliveries.
2. Payment schedule. What triggers each payment, and the dates.
3. Term and termination. How long the partnership runs and who can end it, with what notice.
4. Exact scope per package. Which systems, channels, date ranges and teams.
5. Exclusions. What never leaves (client files, personnel records, anything under an NDA), named in the contract.
6. Exclusivity. Whether you can license the same records to another buyer.
7. End buyers and permitted uses. Which AI companies receive the data, what they may do with it, and whether it can be resold.
8. Access and deletion. What Scale's tooling can reach in your environment, how access ends, and what happens to delivered copies at the end.
9. Warranties and liability. What you promise about your right to license the data, and the cap on your liability.

A range of $10K to $1M+ is too wide to negotiate against on its own. The strongest position is more than one written offer for the same scope, because buyers price the same records differently and their terms differ as much as their prices. Getting competing offers and negotiating them is the work a sell-side advisor does.

If you run a law firm or an accounting firm

Scale lists law firms, legal and compliance, financial services, and business and professional services among its industries of interest (source: Scale AI, data partnerships). Professional rules narrow what those firms can offer. A law firm can never license client files, matter content or privileged communications, because ABA Model Rule 1.6 protects all information relating to the representation of a client (source: ABA Model Rule 1.6). An accounting firm can never license client books or anything derived from a tax return, because IRC 7216 makes unconsented disclosure of tax return information a crime (source: 26 U.S.C. 7216).

What can qualify is firm-owned material with no client information in it: internal procedures, training curricula, staffing and intake workflows stripped of client identity, and templates written from scratch. Our guides for law firms licensing data to AI and accounting firms and IRC 7216 cover what can qualify.

Scale AI referral bonus: $10,000 or up to $50,000

Scale pays people who introduce companies to the program, but its own page gives two different amounts:

• Its FAQ says: “Refer a company and earn up to $50,000 bonus.” Referrals must go through the official referral process and are paid only if Scale signs a partnership agreement. Repeat referrals or introductions to the same company are not paid, and only one referral is paid per deal, to whoever successfully introduces Scale to the right person (source: Scale AI, data partnerships).
• Its “Refer a team” form says: “Refer a company and earn $10,000. Referral bonuses are paid out when Scale signs a partnership agreement” (source: Scale AI, data partnership referral form (page script)).

Both figures were on the page on October 15 and again on November 8, 2026. Scale publishes no written referral terms, so nothing says which amount governs, whether $50,000 depends on deal size, or when the bonus is paid after signing. If you plan to refer a company, get the amount and the trigger confirmed in writing first.

If someone referred you to Scale, they may be paid when you sign. That is common across these programs. Our comparison of AI data referral programs shows what each program pays introducers.

Scale AI vs micro1, Mercor and Handshake AI

Scale sets the highest published size bar and the lowest advertised floor, and it is the only one of the four built around recurring deliveries. micro1 publishes the most detail on pay and referrals, Mercor the fastest payment, and Handshake AI the lowest size bar. Read our profiles of the micro1 data partnership program, the Mercor data program and Handshake AI data partnerships, or the side by side in micro1 vs Mercor vs Handshake. The full list of AI data buyers covers every program we track.

Frequently asked questions

Sources

Program terms checked against each company's own pages on November 8, 2026.

1. Scale AI, data partnerships
2. Scale AI, data partnership referral form (page script)
3. Scale AI blog, new CEO
4. Wikipedia, Scale AI
5. TechCrunch, Jun 13, 2025
6. Scale AI, about
7. TechCrunch, May 21, 2024
8. Fortune, May 21, 2024
9. Fortune, Jun 13, 2025
10. Fortune, Jun 22, 2025
11. ABA Model Rule 1.6
12. 26 U.S.C. 7216
13. micro1, data partnerships
14. micro1, company referral
15. Mercor, Monetize your data
16. Mercor referral app
17. Handshake AI, data partnerships
18. Handshake AI, license your data
19. Handshake AI, refer a company