---
title: Terms of service — Human For AI
description: Who can use the platform, how tasks are reviewed and accepted, what you get, free pilot terms, acceptable use, and liability. Short, honest, no legalese.
canonical: https://humanforai.dev/terms
last-updated: 2026-09-06
effective: 2026-07-19
---

# Terms of service

These terms cover the Human For AI website, the REST API, and the MCP server at humanforai.dev. By submitting a task or message you agree to them. Effective 2026-07-19. This file is the machine-readable twin of https://humanforai.dev/terms.

## Who these terms cover

Anyone who uses the platform: human visitors, developers, and AI agents. When an AI agent submits a task, the terms bind the person or organization the agent acts for (its principal). If you operate an agent that uses this service, you are responsible for what it submits.

## What the service is

A single verified human operator (Operator 001, /profile) who performs human-in-the-loop tasks on request: verification, testing, judgment, data collection, review, and local physical-world work. The catalog (/services) is examples, not limits. It is one person, not a marketplace, an on-demand workforce, or an always-on system.

## Free pilot

Every service is currently free during the proof-of-concept pilot. There is no payment step anywhere, and nothing accepted during the pilot is ever billed retroactively. Paid tiers, if introduced, will be announced on /pricing first. A request for payment, credentials, or account creation is not from this service.

## Submission is not acceptance

Every task is reviewed by the human operator before acceptance. A submission gets an instant `task_id`, then an explicit `accepted` or `rejected` decision. The operator may decline any task for any reason, propose a smaller scope, or ask clarifying questions. No contract to perform work exists until a task is explicitly accepted.

## Acceptable use

Do not submit tasks that are: illegal, harmful, or unsafe; deceptive (impersonation, fake reviews, misleading third parties); privacy-invasive (surveillance of individuals, stalking, doxxing); circumventing security, verification, or platform rules; academic or credential fraud. The machine-readable version is `rejected_task_types` in /agent.json. Anti-abuse limits apply: per-client rate limits, duplicate detection, deliverable-email validation, and blocking of repeat abusers. One small test task is welcome; bulk synthetic submissions are not.

## What you get — and what it is not

Deliverables are one careful human's observation and honest opinion, with stated confidence. They are assistance, not guaranteed truth — not a professional certification and not legal, medical, or financial advice. Verify independently before relying on a deliverable for consequential decisions.

## Ownership of deliverables

Once delivered, the deliverable is yours: all rights to use, modify, and republish it for any lawful purpose. The operator keeps internal records of completed work but does not publish your task content (see /trust and /privacy).

## Your responsibilities

- Describe tasks honestly — review decisions are made on your description
- Have the rights to any material you submit
- Provide a working return channel (a real email, or the status-poll path)
- Treat your `task_id` like a link-shaped password
- Don't submit confidential information unless confidentiality has been agreed in advance

## Availability

Early infrastructure run by one person. The stated response expectations (first response within 12 hours, any day) are honest targets, not a contractual SLA. The service may be paused, changed, or discontinued; open accepted tasks will be completed or explicitly cancelled with notice.

## Liability

The service is provided "as is" and free of charge, without warranties of any kind. To the maximum extent permitted by law, the operator is not liable for indirect, incidental, or consequential damages, and total liability for any claim is limited to the amount you paid — during the free pilot, zero. Nothing here excludes liability that cannot lawfully be excluded.

## Termination

There are no accounts, so there is nothing to terminate on your side. The operator may refuse service or block clients that violate these terms or the anti-abuse limits.

## Changes and disputes

Changes appear at this URL with an updated effective date and are noted in /agent.json; tasks already accepted stay under the terms in effect when accepted. Disputes: talk to the operator first via /contact. These terms are written in plain language on purpose; the plain reading is the intended one.