Handshake AI vs Mercor: How They Compare for AI Data Work (2026)

At a glance#

Figures are as of October 2026. Revenue numbers are company claims and unaudited.

Why this comparison matters#

Handshake AI and Mercor are two prominent names in the expert-data market, and they reached it from related places: Handshake from campus recruiting and Mercor from freelance matching. Handshake built a campus careers network. Mercor started by matching freelance engineers with companies.¹¹ When AI labs began paying for expert judgment from lawyers, scientists and engineers, both already had people and employer relationships to draw on.

Both also publicly run data-licensing programs aimed at ordinary businesses: besides supplying expert labor, each asks companies to contribute their own operational records.¹⁵ For a business owner, that makes this a directly relevant comparison, and also one where the claims on each vendor's page deserve scrutiny.

This page covers what each does, how it is funded, how it sources and pays people, what disputes are on the public record, and what the data programs say publicly. It avoids anything we could not tie to a source.

Handshake AI#

Background#

Handshake was founded by Garrett Lord, Ben Christensen and Scott Ringwelski, Michigan Tech students of the class of 2014. Its first client school was Wabash College, and by 2024 the company reported more than 20 million students and alumni, 1,600 universities and 1 million employers.¹⁰ Handshake was last valued at $3.5 billion in January 2022, before its move into AI training, according to Business Insider.⁸

The AI pivot#

Handshake AI was conceived at a 2024 Christmas party, per Michigan Tech's alumni magazine.¹⁰ In October 2025 Lord described to Upstarts Media a "refounding" of the company around AI; the company had laid off about 100 people, 15 percent of its workforce, and its AI team had grown from 15 to 150 people and reached a $100 million annualized run rate in about eight months.¹⁷

TIME reported in 2026 that Lord says the AI unit passed $1 billion in gross annualized revenue about 16 months after launch, works with eight frontier AI labs that it does not name, and draws on hundreds of thousands of experts.⁶ The $1 billion figure is the CEO's own claim and is not audited, and we found no public figure for net revenue after contractor payments.

Products and business model#

• Handshake AI Fellowship. Remote, part-time projects where contributors train and evaluate models. The page cites $500 million or more in payouts, 188,000 or more fellows, and a headline rate of up to $100 an hour. Contributors must be based in the US with work authorization Handshake can support.²
• Enterprise and research offerings. The AI page links to Enterprise AI and Research sections without describing them in detail.²
• Data partnerships. A separate program that licenses business data to AI labs (covered below).³

Contributor pay and terms#

The fellowship page lists rates from up to $17 an hour (AI evaluation specialist) to up to $125 an hour (game developer, PCB tool specialist).² Business Insider reported listed freelance rates from $75 an hour for software engineers and improv actors to $175 for investment bankers and $300 or more for holders of medical degrees or PhDs, and said Lord had stated on a July podcast that contractors average $100 to $125 an hour.⁸

Handshake's Contractor Agreement, last updated September 17, 2026, states that contractors are independent business enterprises; contractors assign all rights in work product to Handshake; use of AI tools to produce work product is banned without written authorization; disputes go to individual arbitration under JAMS rules with a class action waiver and a 30-day opt-out for new contractors; Handshake can terminate on seven days' notice and may deactivate accounts temporarily without notice. It also provides for background checks and remote monitoring.¹⁸

Disputes#

Business Insider reported in March 2026 that five contractors on OpenAI projects said Handshake suspended their accounts without warning between late December and January, and that four were not paid for their work. One was told they had broken platform rules involving credential discrepancies, task times three to four times the benchmark, and work from outside the US, which the contractor denied. Handshake support told that contractor the decision was final with no appeal process, and a Handshake representative declined to comment. The article also described a wage suit in which a court ruled in October that a contractor should be paid $6,475, and a November suit seeking $9,600 that the contractor moved to dismiss in mid-January.⁸ OpenAI did not respond to requests for comment.⁸

Mercor#

Background#

Mercor was founded in 2023 by Brendan Foody, Adarsh Hiremath and Surya Midha, friends from a Bay Area high school debate team who are all Thiel Fellows, according to Forbes. Forbes reports they started by matching engineers in India with US companies that needed freelance coders.¹¹ TechCrunch reports that Mercor now supplies domain experts for AI model training, charging an hourly finder's fee and matching rate.⁹

Funding and growth#

On October 27, 2025, Mercor confirmed a $350 million Series C at a $10 billion valuation, led by Felicis, with Benchmark and General Catalyst participating and Robinhood Ventures joining. The prior round was $100 million at $2 billion.⁹ TechCrunch reported at the time that Mercor had more than 30,000 experts, average pay above $85 an hour and daily contractor payouts above $1.5 million, and that Mercor was on track to hit $500 million in annual recurring revenue.⁹

In July 2026 TechCrunch reported Mercor was in early talks to raise at a $20 billion valuation, per Bloomberg, and that CEO Brendan Foody said on X that its annualized revenue run rate had crossed $2 billion, double from four months earlier.¹² That is the CEO's statement, not an audited figure, and the report does not say whether it is gross or net of contractor payments. We found no report that the $20 billion round has closed. Mercor also announced it was buying Deeptune, a company that helps train AI agents.¹²

Products and business model#

Mercor's homepage lists three services: human data and evals, data monetization (licensing anonymized workflow data that businesses generate), and enterprise AI (mapping workflows and deploying AI with embedded engineers). It also publishes the APEX benchmark family and names Harvey, Ramp, Cognition and Sierra as benchmark partners, which it does not describe as paying customers.⁴ Mercor Enterprise describes four stages (Discover, Deploy, Improve, Monetize), names no customers, and gives no pricing.¹⁹

Contributor pay and terms#

Mercor's homepage claims a $107 average hourly rate, more than $4 million in daily payouts, over 30,000 experts, and role rates from $60 to $70 an hour for a sales and marketing expert up to $350 to $450 an hour for an oncologist.⁴ CBS News's transcript of a 60 Minutes segment dated October 4, 2026 said Mercor has assembled more than 100,000 freelancers and that typical trainers are paid $20 to $200 an hour.⁷ In May 2026 CBS reported that Mercor said its AI training jobs pay an average of $105 an hour.²⁰ These numbers differ by source and date, and the platform's own figures are self-reported.

Mercor's general Terms of Service say experts are not employees of Mercor and that Mercor "does not control the manner or means of the services provided by the Experts," and refer arbitration terms, and in some cases ownership, to a separate expert contractor agreement we could not review. They also grant Mercor a broad license over user content, including for training models.²¹

Disputes, breach and lawsuits#

• Scale AI suit (2025). In September 2025, Scale AI sued Mercor and Eugene Ling, a former Scale employee, in federal court in Northern California, alleging he copied more than 100 confidential documents to a personal Google Drive and pitched a Scale client on Mercor's services. Mercor co-founder Surya Midha denied any misuse and said Mercor had "no interest in any of Scale's trade secrets." We found no reported ruling.²²
• Security incident (2026). TechCrunch reported that on March 31, 2026 Mercor admitted a data breach it attributed to a compromise of the open-source LiteLLM tool, which harbored credential-harvesting malware for about 40 minutes. A hacker group claimed 4 terabytes of stolen data; Mercor had not commented on the claim's authenticity. Meta paused its contracts with Mercor indefinitely, per sources cited by Wired, and OpenAI confirmed to Wired that it was investigating its exposure.¹⁵ In a June 25, 2026 update, Mercor said the compromised LiteLLM versions affected a significant number of companies globally, that its investigation was complete, and that only a very limited subset of its nearly five million experts had sensitive information affected, with no evidence of fraudulent use. It said it was notifying affected individuals and had shared findings with customers.¹⁶
• Class actions. PYMNTS, citing the Wall Street Journal, reported on April 23, 2026 that Mercor had been served with at least seven class actions alleging exposure of contractor information such as interview recordings and facial biometric data, and further alleging that Mercor monitors contractors' computers, uses recorded interviews to train models, and trains client models on materials possibly owned by other companies. Mercor told the Wall Street Journal, "We strongly dispute the speculative claims in these lawsuits."²³ One of the suits, Ananthula v. Mercor.io Corporation, was filed on April 21, 2026 in the Northern District of California by the law firms Hausfeld and Hall Attorneys; it alleges failure to implement reasonable data security measures and negligence. These are allegations, not findings.²⁴

Head to head#

Scale and money#

Mercor's CEO claims about twice Handshake AI's run rate (above $2 billion in July 2026 versus above $1 billion), and Mercor's valuation is a priced $10 billion round with $20 billion talks reported.¹²⁶⁹ Handshake's last priced round we found was in 2022 and predates its AI business.⁸ Handshake's figure is described as gross; the TechCrunch report does not say whether Mercor's is gross or net. Neither company's net revenue or margins are public, and neither figure is audited.

Origins and network#

Handshake draws on an existing network of 30 million users, 1 million employers and 1,600 schools, which TIME names as a distribution advantage.⁶ Mercor was founded in 2023 and began by matching freelance engineers with companies.¹¹ Handshake's fellowship is limited to US-based contributors, while Mercor's terms and pages do not state a country limit in the pages we reviewed.²⁴

Pay and terms#

Handshake's CEO has said contractors average $100 to $125 an hour, while its fellowship page posts "up to" rates by role; Mercor's site claims a $107 average and lists role rates with higher top-end specialties.⁸²⁴ Both treat experts as independent contractors, with Handshake's agreement requiring individual arbitration and Mercor's terms referring arbitration to a separate contractor agreement, and both have public worker disputes of different kinds: withheld pay at Handshake, and breach-related and working-condition claims at Mercor.^8182123

Enterprise and data licensing#

This is where the two are closest. Handshake's data partnership page lists sectors including finance and insurance, legal, healthcare, manufacturing, retail and software, and says you can see what data you are licensing, how it can be used and how payment works before committing.¹ Mercor's data page lists similar sectors and describes four steps: scope, connect, process, payout.⁵ Handshake publishes a payout range and a minimum company size; Mercor does not publish a rate card.¹³⁵

Timeline of the public record#

The dates below are the ones that matter for judging how new and how fast-moving these programs are.

• 2014: Handshake's founders graduate from Michigan Tech; Wabash College is its first client school.¹⁰
• January 2022: Handshake is valued at $3.5 billion before its AI pivot.⁸
• 2023: Mercor is founded.¹¹
• September 2025: Scale AI sues Mercor over alleged trade-secret misuse.²²
• October 2025: Mercor raises $350 million at a $10 billion valuation; Handshake announces its AI refounding and layoffs in the recruiting business.⁹¹⁷
• March 2026: Business Insider reports the Handshake contractor pay disputes.⁸
• March 31, 2026: Mercor confirms it was affected by the LiteLLM compromise.¹⁵
• April 2026: Mercor is served with at least seven class actions.²³
• June 25, 2026: Mercor publishes its security incident update.¹⁶
• July 2026: Mercor is reported to be in talks at a $20 billion valuation.¹²
• October 2026: 60 Minutes covers Mercor's freelancers; Handshake's data partnerships page is dated October 20.⁷¹

Strengths and weaknesses#

Handshake AI#

• Strengths: a very large pre-existing network, fast claimed growth, a published payout range and eligibility for data partners, and a standard contractor agreement that is public.⁶¹³¹⁸
• Weaknesses: the revenue claim rests on the CEO's statement, contractor suspension and pay disputes reported by Business Insider, and no named labs or buyers.⁶⁸¹

Mercor#

• Strengths: a higher claimed run rate, a well-funded balance sheet, a visible benchmark research program and an opt-in data licensing flow in which, it says, no data leaves your systems until you approve it.¹²⁹⁴⁵
• Weaknesses: the 2026 breach and resulting suits, the Scale AI litigation, customer pauses reported by Wired through TechCrunch, and no published data pricing.¹⁵²²²³

Who each suits#

Handshake AI suits US-based experts who want part-time fellowship work with rates posted upfront, and businesses with 20 or more employees and three or more years of records that want to see an estimate under a published range.²¹³ Mercor suits experts across many fields who are comfortable with a large marketplace, and enterprises interested in its human data, benchmarks or agent deployment services.⁴¹⁹ Anyone who plans to connect systems to Mercor should review its security disclosures first.¹⁵

What this means for data sellers#

The data-seller question has a direct, documented answer here: yes, both companies run programs to license a business's operational data to AI labs.³⁵ That is useful market evidence of demand for operational records from sectors such as legal, healthcare, finance, manufacturing, retail and logistics, which both vendors list.¹⁵

The public pages also leave important questions open.

1. Who the buyers are. Both pages refer only to "AI labs" and name none.⁵³
2. How de-identification works. Handshake says it will confirm "any required de-identification" and who prepares the data; Mercor says extraction and de-identification are automated and that buyers never see raw data. Neither describes its method in detail on the pages we reviewed.¹⁵
3. What the payout depends on. Handshake's estimate is "indicative only" and "not an offer to purchase," and final payout depends on data sources, volume, quality, due diligence and acceptance of licensing terms.¹⁴ Mercor says pay depends on volume, tools connected and depth.⁵
4. Exclusivity, retention and term. Not stated publicly by either.³¹⁴
5. Security track record. Mercor's 2026 incident is the main reason to ask for security documentation and audit logs before connecting any tool.¹⁵
6. Referral incentives. Handshake pays a $10,000 bonus to referrers when a deal closes, and Mercor pays a bonus when a referred company becomes a paid partner, so some third-party recommendations carry a financial interest.¹³⁵

If you are weighing this route, start with what data you have, read how selling works and our trust notes, and consider getting independent legal review of any license. SellYourAIData is not affiliated with Handshake or Mercor, and nothing on this page is an offer from either.

Do not let one buyer set your price. Any single buyer's published range, program page or first offer is that buyer's number, not the market's. The sensible move is to get competing offers on the same data, compare scope, exclusivity, indemnity and payment timing side by side, and have someone independent look at the terms. Our guide to selling direct vs using a broker explains how, and if a buyer is already talking to you, we review offers free.

Questions to ask before you share anything#

Neither vendor's public page replaces a contract. If you are a company considering either program, these are reasonable questions to put in writing:

1. Who are the intended buyers or categories of buyer, and can any be excluded?
2. Where does de-identification happen, and does any raw copy of your data reach the vendor first?
3. Is the license exclusive or non-exclusive, for what term, and for which uses (training, evaluation, resale)?
4. What is retained after delivery, and what happens if you terminate?
5. Which of your own customers, employees or partners appear in the records, and do your existing contracts allow licensing at all?
6. How and when is payment made, and what happens if due diligence reduces the scope?
7. What security certifications and incident history can the vendor document?

Handshake says you can see what data you are licensing, how it can be used and how payment works before committing, and Mercor says no data leaves your systems until you approve it.¹⁵ Those are statements of intent on public pages. Hold the vendor to them in the agreement.

What we could not verify#

• Handshake's $1 billion figure comes from its CEO via TIME and is not independently verified.⁶
• Mercor's $2 billion figure is its CEO's statement as reported by TechCrunch; we do not know whether it is gross or net, and it is unaudited.¹²
• We could not read Handshake's data licensing terms or Mercor's expert contractor agreement.
• Mercor's class actions are allegations; the count (at least seven) comes from the Wall Street Journal as relayed by PYMNTS and may have changed.²³
• We did not find a stated founding year for Handshake in the sources we cite; its founders were Michigan Tech's class of 2014.¹⁰

Related reading#

For broader background, see our guide to selling data to AI companies and the Mercor vs micro1 comparison.