Trust centerContact sales
Start free
Products
Proof-of-Human APIKnow Your AgentData and evalsVerified human trafficHuman tasksVerified clippingPricing
Developers
DocsVerify a sealAI agentsRegister an agentMCP server
People
How it worksGet verifiedYour dashboard
Learn
Learn hubProof of humanKnow Your HumanAgentic commerceCAPTCHA alternativesComparisons
Company
AboutTrust centerContactHelp
Sign in
Start freeGet verified — free
← All products
Data and evalsEarly access

Ground-truth data, signed row by row.

RLHF rankings, annotation, moderation and transcription from ID-verified people. Every row carries a seal your teams can check offline.

Talk to us How it works
  • ID-verified labelers
  • A seal on every row
  • Pay per accepted label

Early access: we set up each project with you directly.

ML researcher reviewing human-labeled training data on two monitors AI-generated image
A researcher reviews labeled data; each example row shows a verified person and a valid seal.
  • RLHF / DPO preferences
  • Safety and moderation
  • Image annotation
  • Translation QA
  • Audio transcription
  • Document understanding
How it works

From spec to signed labels.

  1. Define the task

    Paste a rubric or start from a template. Set the label schema and what counts as accepted.

  2. Set rate and volume

    You only pay for labels you accept.

  3. Verified people label

    ID-verified contributors do the work. Each label is sealed to the live person who made it.

  4. Export with provenance

    Rows arrive in your schema — CSV, JSON or webhook — each with its seal.

What you get

Labels you can prove a human made.

Most vendors ask you to trust the labels. We hand you proof you can check yourself.

Verified people only

Every contributor passes ID and liveness first. The same accountable people stay in the pool, not an anonymous crowd.

Proof on every row

Each row carries a seal: anonymized contributor, timestamp, task, and a consensus score where you need agreement.

Bot-resistant by design

A bot can’t produce a live person’s signature. Change one character of a row and it fails verification.

The data you receive

Every label ships with its proof.

A record in your schema, ready for ML, audit and procurement.

  • The label itself, in your schema
  • Verified, anonymized contributor reference
  • Timestamp and the task performed
  • ES256 signature you can check offline
  • Consensus score where you need agreement
export.jsonSample
// one label in your export
{
  "label": "safe",
  "task": "content_safety",
  "contributor": "vh_4821", // identity-verified
  "verified": true,
  "live": true,
  "consensus": 0.96,
  "seal": "eyJhbGciOiJFUzI1NiIs…", // ES256
  "ts": "2026-06-25T14:02:11Z"
}
Who it’s for

For teams that need defensible data.

Woman comparing two answers on her laptop at a kitchen tableAI-generated image
AI labs

RLHF and DPO data

Pairwise rankings with provenance, so alignment data is auditable.

Two colleagues reviewing labels on a laptopAI-generated image
Trust and safety

Moderation labels

Policy-graded safety labels from accountable people, with a proof trail.

Man tagging photos on a tablet on his porchAI-generated image
Computer vision

Image annotation

Boxes, masks and hard edge cases, consensus-scored where needed.

Student with headphones translating in a libraryAI-generated image
Multilingual AI

Translation and documents

Native-speaker review and structured document extraction.

Pricing

Pay for accepted labels.

No seat licenses, no lock-in. Checking the proof on any label is always free.

Pay-as-you-go
Per label
  • No minimum
  • All data types
  • Proof-of-work export
  • Email support
Talk to us
Recurring pipelines
Volume rates
  • Custom templates and API
  • Webhook callbacks and SLA
  • Provenance dashboard
Talk to us
Enterprise
Custom
  • Prepaid volume blocks
  • MSA, DPA, data residency
  • Dedicated verified cohorts
Contact sales

Data and evals is in early access, so we set up each project with you directly. First-batch pilots are available for AI labs and academic research groups — ask about a free pilot. Want the dashboard and history for everything you seal? Plans from $99/mo.

FAQ

Questions data buyers ask.

What data types can I order?

RLHF and DPO preference data, safety and moderation, image annotation, translation QA, audio transcription and document understanding. If a person can judge it, you can order it.

How do I know a real human produced each label?

Every contributor passes ID and liveness through our verification partners and is live when they label. Each row carries a seal anyone on your team can check offline against our public key.

How is pricing structured?

You pay per accepted label, with volume rates for recurring pipelines and prepaid blocks for high annual volume. Checking the seals is always free.

How fast are first results?

It depends on volume and task type; small batches can come back within hours of a spec. Connect the API and webhooks to stream larger volumes.

What stops bots or click farms?

Every label must be signed by a live, ID-verified person, and a bot can’t produce that. A copied or edited row fails verification.

Can I get a provenance export for compliance?

Yes. Every batch includes the seal for each row, so procurement and audit can verify it offline, against our public key. 2 patents pending (US provisional).

What does early access mean?

Data and evals is open, but not self-serve yet. We scope and set up each project with you directly.

Ground-truth data, with proof.

Send us your first spec. Every row comes back signed.

Talk to us All products