fairdock.

Fairdock · Hong Kong

The data layer for world models.

We source, clear and license the data that world models and embodied AI learn from: gameplay with inputs, first-person video, robot teleoperation, 3D assets and game environments. From holders worldwide to AI labs across Asia and beyond.

We have direct relationships in China’s AI industry, including with suppliers to its leading labs.

Request data License your data

Provenance on every file / Non-exclusive / Compliance built in

How data movesIllustrative
  1. Studios, collectors, robot operators
  2. fairdock.Hong Kong
  3. AI labs building world models

Example dataset card

Origin:
AAA and indie titles, players worldwide
Includes:
60 fps video · keyboard, mouse and controller · game events
Checks:
Inputs aligned to video · Usernames blurred · Publisher terms checked

What we license

Gameplay with inputs

First-person video

Robot teleoperation

3D assets & game environments

Company records: email, chat, docs, code

Motion capture

And more

01 / For AI labs

Training data for
world models and agents.

Tell us the data you need. We source it from studios, collectors and robot operators worldwide, and deliver it with the licence chain and provenance records your legal team expects.

The specification:
modality, volume, format, labels and timeline.
The sample:
reviewed before you commit to anything.
The records:
licence chain, consent basis and content screening results.

支持中文沟通。Chinese-language support.

Share your data requirements

02 / For data holders

Your data, in the
training sets that matter.

Game studios, data collectors, robot operators and 3D artists hold what AI labs are buying. We find the buyer, handle the paperwork and the Chinese-language side, and run the deal end to end.

Non-exclusive:
your existing buyers and channels stay yours.
You set the floor:
per hour, per asset or per dataset, and we sell at or above it.
Paid per deal:
payment before full delivery.
Tell us what you hold

Closing, restructuring or short of cash? AI labs also license a company’s operating records: email, chat, documents, tickets, CRM and code. Tell us what you hold

How it works

Source. Sample.
License.

Every deal starts with a real specification and a sample, so both sides know exactly what they are getting before money moves.

  1. 01

    Tell us what you need, or what you hold

    AI labs share the specification: modality, volume, format, labels and timeline. Data holders share what they have, how much, and how the rights were obtained.

  2. 02

    We source and clear

    We map each request to holders whose data fits, confirm the licence chain and consent basis, and screen every buyer and its owners against sanctions and restricted-party lists.

  3. 03

    Sample first

    The lab evaluates a sample before committing. No-one licenses blind.

  4. 04

    License and deliver

    A licence sets the use, territory, term and price. Full delivery follows payment, and the buyer deletes the data at the end of the term.

Rights and compliance

Clean data,
clean paperwork.

Provenance on every dataset

Licence chain, consent basis and content screening, recorded for every dataset we deliver.

Every buyer screened

Buyers and their owners are checked against US, EU, UK and Canadian lists before any sample is shared.

No sensitive US-person data for China

For China-bound deals we exclude the categories US rules restrict, including biometrics, precise location, health and financial data.

Non-exclusive by default

Holders keep their existing buyers and channels. We license non-exclusively, with the right to resell across Asia.

Questions

Before we get started.

What data do you work with?

The data world models and embodied AI learn from: gameplay recordings with synced keyboard, mouse and controller inputs; first-person video of real tasks; robot teleoperation episodes with camera, joint-state and action streams; 3D assets, scenes and game environments; motion capture; and simulation or synthetic data. It has to exist already and be processed.

Who are the buyers?

AI labs and data teams building world models, embodied AI and agents. Most are in Asia, including mainland China, Hong Kong and Singapore, and we work with labs elsewhere too.

Is it legal to license our data into China?

For licensed content without personal data, generally yes. US rules restrict selling Americans’ sensitive personal data to China-linked buyers, so we exclude those categories for China-bound deals. Your own counsel signs off on every licence.

Is the arrangement exclusive?

No. Licences are non-exclusive by default, and your existing buyers and channels stay yours.

How is pricing set?

Holders set a floor price per hour, per asset or per dataset. We agree the final price with the buyer at or above that floor, and holders are paid before full delivery.

Do you buy the records of closed companies?

Yes. AI labs license the day-to-day operating records of companies that have closed, are winding down or are in liquidation: email, chat, documents, tickets, CRM, code and call recordings, with personal data removed where required. Founders, directors and liquidators can tell us what they hold at fairdock.ai/sell.

Can you deliver in a specific format?

Yes. We deliver in the structure the lab specifies, from raw recordings with aligned input logs to packaged tasks with verifiers, and we document the provenance of every file.

Meet the founder

Ningnan Ma

Founder of Fairdock

Connect on LinkedIn

The vision

World models will be trained on the real world’s data. Fairdock exists to get that data to the labs building them, with the rights in order.

Let’s talk

Tell us what you need,
or what you hold.

AI labs: share the specification and we’ll come back with what we can source and a sample. Data holders: share what you have, how much, and how it was collected.

Or email us at admin@fairdock.ai.

I’m reaching out as a

We’ll use these details to respond to your enquiry, as described in our privacy notice. Please don’t include sensitive data or confidential records.