DataOcean AI is the international brand carried by DATAOCEAN AI INC. and DATAOCEAN AI PTE.
LTD., subsidiaries of Beijing Haitian Ruisheng Science Technology Ltd., a Chinese AI training data company founded in 2005 and listed in Shanghai as 688787; it sells speech, image and text datasets and custom data projects, and no program for buying data from outside holders is on record for it.
This reference on AI training data licensing covers every buyer and vendor on the same dated-source method.
Filed figures come from the parent's reports in CNY, interim and unaudited, read 2026-11-16; website figures are DataOcean AI's own claims, checked 2026-11-09. Its 2026 half-year report states no major litigation or arbitration, and no route or terms for data holders are on record as of 2026-11-16.
What DataOcean AI Sells: Off-the-Shelf Datasets and Custom Data Projects
DataOcean AI sells AI training data rather than buying it: its parent's 2026 half-year report describes custom datasets made to order and a catalog of standard dataset products licensed to customers.
The same report lists data tools and model training and evaluation services next to the datasets. The data types are speech, vision, text and multimodal material, and the parent describes speech recognition and speech synthesis data as its long-standing line.
The size of the catalog appears in three numbers. The website states "over 1800" and "over 1600" off-the-shelf datasets on the same page, checked 2026-11-09, and the half-year report counts 1,912 training data products with the company's own intellectual property at 2026-06-30, with more than 35 added in the first half of 2026.
The counts differ in kind and date, so they are not merged. DataOcean AI is a seller to AI developers and is not among who buys AI training data from outside holders.
DataOcean AI Company Identity: Haitian Ruisheng and the DATAOCEAN AI Subsidiaries
The company behind DataOcean AI is Beijing Haitian Ruisheng Science Technology Ltd., listed on the Shanghai Stock Exchange STAR Market under stock code 688787 since 2021-08-13, whose filings list two subsidiaries named DATAOCEAN AI.
The table lists the identity facts and the type of evidence behind each, as of 2026-11-16.
| Attribute | Value | Evidence type | Source |
|---|---|---|---|
| Brand | DataOcean AI | Website | dataoceanai.com |
| Website | dataoceanai.com | Website | dataoceanai.com |
| Website operator as stated | "DATAOCEAN AI", with no legal form, address or country | Privacy policy and footer | dataoceanai.com/privacy-policy/ |
| Legal entity operating the website | Not stated | n/a | n/a |
| Listed parent | Beijing Haitian Ruisheng Science Technology Ltd. (北京海天瑞声科技股份有限公司), SSE STAR Market 688787, listed 2021-08-13 | Exchange company profile | sse.com.cn |
| Registered office | Haidian District, Beijing | Exchange company profile | sse.com.cn |
| Subsidiaries named DATAOCEAN AI | DATAOCEAN AI INC., Washington State, United States, 100 %; DATAOCEAN AI PTE. LTD., Singapore, 75 % (51 % at 2025-12-31) | 2026 half-year report | static.sse.com.cn, 2026-08-27 |
| Founded | 2005 | Company's own claim, checked 2026-11-09 | dataoceanai.com/about-us/ |
| Link to the brand | Parent's contact page gives www.dataoceanai.com and contact@dataoceanai.com; speechocean.com addresses appear only in its page source | Parent's Chinese website | haitianruisheng.com |
| Executives and headcount | Not stated on this page | n/a | n/a |
| Type | Speech and multimodal off-the-shelf dataset vendor | Description | dataoceanai.com |
The exchange profile, the 2025 annual report and the 2026 half-year report tie the names together: the parent files the DATAOCEAN AI subsidiary lists, and its contact page names dataoceanai.com as its website.
Older contact, supplier and recruitment addresses at the speechocean.com domain remain in that site's page source, which ties SpeechOcean to the same group; no source read here says one name replaced the other.
The website names its operator only as "DATAOCEAN AI", and the filings do not say which group company runs it. The contracting entity of any deal is therefore fixed in the contract itself, not inferred from the brand.
DataOcean AI Terms for Data Holders, Dated
No program, form or terms for buying or licensing data from outside holders are on record for DataOcean AI; the table lists each term a holder would check and what the record holds, as of 2026-11-16.
| Term | Stated value | Checked | Source |
|---|---|---|---|
| Program to buy or license data from outside holders | None on record | 2026-11-16 | dataoceanai.com |
| License scope, term, exclusivity, deletion for holder data | None published | 2026-11-16 | dataoceanai.com |
| Payout to data holders | None stated | 2026-11-16 | dataoceanai.com |
| Payment trigger and timing | None stated | 2026-11-16 | dataoceanai.com |
| Company referral fee | None found | 2026-11-16 | dataoceanai.com |
| Speaker and contributor pay rates | Not recorded | 2026-11-16 | no record |
| Speaker consent form and contract terms | Not recorded | 2026-11-16 | no record |
| Supplier route for collection and annotation work | Exists per the parent's report; terms not published | 2026-11-16 | 2026 half-year report |
Every holder row reads "none", so a holder has no published scope of use, term, exclusivity rule, deletion duty, price or payment trigger to weigh. Any deal would be negotiated from nothing.
The two contributor rows read "not recorded" rather than "none". The filing describes recruited speakers and paid suppliers, so such terms exist, but they are not published in the sources checked.
Programs that publish company terms are set side by side as program terms compared. Read beside them, this table is a list of gaps: the price, timing and rights columns that those programs fill are empty here.
How DataOcean AI Sources Speech and Other Training Data
DataOcean AI's speech data is made rather than bought in bulk: its parent's half-year report says the company recruits speakers to read and record scripts it designs, then labels and checks the audio, or processes audio that clients supply.
The report describes the steps as designing the scripts, scenes and speaker mix, recruiting speakers, recording, segmenting the audio with time stamps, labeling it and running a quality check. Client-supplied audio skips the recording step and enters at processing.
Image, video and text data follow the same pattern: organized photo and video capture, collected or written text, or client files, then labeling and checks.
Much of the raw collection and labeling is bought in. The report calls it data service procurement, and data service fees were 84.29 % of main-business-related procurement in the first half of 2026.
It also states a supplier network of 650,000 terminal resources across 178 countries and regions and more than 300 languages or dialects, while the website claims collection resources in more than 70 countries (checked 2026-11-16). Both are claims. How the market for such recordings works for sellers is covered under selling speech and audio data.
Not on record: what speakers are paid, the contracts and consent forms they sign, and whether any licensed-in third-party dataset enters the catalog.
Who Owns the Data: DataOcean AI Custom Projects Versus Catalog Products
Ownership depends on the product line: in a custom project the client pays a service fee and the company says it holds no IP in the data and cannot resell it, while catalog datasets are its own IP and are licensed to many customers.
The half-year report describes the catalog model as one-time development and repeated sale. The license a customer receives for a catalog dataset is not on record, and neither is its price. How such terms usually differ is set out under dataset licenses for AI training.
For a holder, the split matters in both directions. Files supplied as a client come back as the client's dataset, not resold under the company's description of custom work. Data that the company collects for its own catalog becomes its asset and can be licensed repeatedly.
DataOcean AI and Chinese Data Law: What Its Filings Name as Risks
Haitian Ruisheng's own filing treats Chinese data law as a business risk, naming the Personal Information Protection Law (in force 2021-11-01) and the rules on sending personal information collected in China abroad.
The half-year report also names the Interim Measures for the Management of Generative Artificial Intelligence Services (in force 2023-08-15) and the Network Data Security Management Regulations (in force 2025-01-01). It adds that overseas business within the scope of the cross-border rules may be affected to some degree if procedures are not completed.
The statutes themselves are explained under China's rules for licensing data, including consent and export duties for data used in AI training. This page is information, not legal advice.
What a Data Holder Can and Cannot Do With DataOcean AI
A company cannot license its data to DataOcean AI through any published route, because no program, form or terms for data holders are on record as of 2026-11-16; it can approach DataOcean AI as a customer or as a supplier of collection work.
An approach made in the name of DataOcean AI or SpeechOcean arrives outside any public program, so it is checked against the domains dataoceanai.com and haitianruisheng.com before anything is sent. The group also states a supplier route, but its terms are not published.
A company can be a customer, buying catalog data, commissioned collection or annotation. Its confidentiality and data-protection duties then travel with any files handed over.
A holder who wants a counterparty with published terms is at a different step: choosing a buyer for your data among labs, vendors and marketplaces.
Haitian Ruisheng's Filed Figures and DataOcean AI's Own Claims
Haitian Ruisheng reported revenue of CNY 290,275,912.04 for the first nine months of 2026, up 23.87 % year on year, in its third-quarter report dated 2026-10-31.
The table shows filed group and subsidiary figures in CNY (interim and unaudited) and the website's claims, as of 2026-11-16.
| Figure | Value | Period or date | Source |
|---|---|---|---|
| Group revenue | CNY 290,275,912.04, +23.87 % | 2026-01-01 to 2026-09-30 | Q3 report, 2026-10-31 |
| Group net profit attributable | CNY 10,460,300.91 | 2026-01-01 to 2026-09-30 | Q3 report |
| Q3 2026 revenue | CNY 98,395,037.88, +26.73 % | Third quarter 2026 | Q3 report |
| H1 2026 revenue | CNY 192 million (as filed), +22.45 % | First half 2026 | Half-year report, 2026-08-27 |
| Overseas share of main-business revenue | 26.65 % | First half 2026 | Half-year report |
| DATAOCEAN AI INC. | Revenue CNY 4,030,000.60; net loss CNY 1,074,876.39 | First half 2026 | Half-year report |
| DATAOCEAN AI PTE. LTD. | Revenue CNY 28,122,557.51; net profit CNY 2,920,557.83 | First half 2026 | Half-year report |
| Own-IP dataset products | 1,912 | At 2026-06-30 | Half-year report |
| Off-the-shelf datasets | DataOcean AI states "over 1800" and "over 1600" on the same page | Checked 2026-11-09 | dataoceanai.com |
| Clients | DataOcean AI states "more than 1100" and "more than 900" on the same page | Checked 2026-11-09 | dataoceanai.com |
Group revenue grew while overseas revenue fell year on year: the overseas share was 26.65 % of main-business revenue in the first half of 2026, and domestic revenue rose 69.18 %.
The Singapore unit was profitable and the US unit loss-making in that half-year. The website pairs are shown as stated, and no market value or client name is given.
DataOcean AI Reputation: The Record as of 2026-11-16
Haitian Ruisheng's 2026 half-year report states that the company had no major litigation or arbitration in the period, and no court case, breach notice or regulator action involving DataOcean AI is on record in the sources checked as of 2026-11-16.
A source here is a court docket, a regulator notice or a company filing or statement. Review sites, job boards, app-store listings and similarly named companies, such as DigitalOcean and Data.AI, are not used. The dated log of data buyer incidents holds no DataOcean AI entry, and new entries go there.
The absence of a record is not a clean record. It means only that nothing was found in those sources.
Is DataOcean AI Legitimate as a Counterparty for Company Data?
DataOcean AI is backed by an identifiable listed company that publishes filings, but it is not a documented buyer of company data, because no route for data holders is on record.
Documented: a listed parent with periodic filings, two named subsidiaries, a founding year of 2005 by its own claim and its own website.
Missing: the legal entity behind the website, any seller program, speaker pay and consent terms. Legitimacy as a company and fit as a counterparty for licensing data are separate questions, and on the second the record is empty.
A message that uses the DataOcean AI or SpeechOcean name is checked by its sender domain and by the contracting entity in the draft. The checks for whether a data licensing email is legitimate apply.
Benefits, Risks, Downsides and Alternatives for a Data Holder Considering DataOcean AI
For a data holder, DataOcean AI is an identifiable vendor to buy datasets or collection work from, not a published route to license data to, so the alternatives are programs that publish terms and the option of not licensing at all.
Benefits
- A listed parent with periodic filings
- A stated rule that custom data belongs to the client and is not resold
- Speech data experience claimed since 2005
Risks
- The website operator is not named, so the contracting entity is fixed in writing
- Catalog data is licensed to many customers
- The filing names China's cross-border transfer rules as a risk
- Approaches in its name cannot be matched to a program
Downsides
- No stated price, payout or timing
- No intake route for data offers
- Speaker terms not on record
Alternatives (among them, alphabetical, with the date each claim was checked)
- Abaka AI Project Reborn: Abaka states up to USD 1M per repo and up to USD 5M per company, paid in two parts (2026-11-09)
- Claru Partnerships: Claru states archive licenses paid on signed license, net-15, with a typical first archive deal of USD 5K to 50K (2026-11-09)
- Defined.ai Data Partners: Defined.ai states Data Partners earn licensing fees from marketplace sales on terms the partner approves, with no revenue share percentage stated (2026-11-09)
- Handshake AI Data Partnerships: Handshake states "Earn $100K to $4M+", while its estimator on the same page reads "up to $100,000" (2026-11-16)
- Labelbox Prism: Labelbox states datasets "typically start around $50K" and enterprise partnerships "can reach $1M+", as an upfront fee plus revenue share (2026-11-09)
- micro1 Company Data Partnerships: micro1 states USD 100K to 500K for qualified enterprise data partnerships (2026-11-09)
- Scale Data Partnerships: Scale states "$10K to $1M+ illustrative value per data partnership" (2026-11-09)
- A direct deal with an AI lab
- Not licensing at all
These are program claims and ranges, not money paid to any holder. Whether to sell or license your data at all comes before any of them.
Where DataOcean AI Sits Among Speech Data Vendors and Data Programs
DataOcean AI is one of the catalog sellers among dataset vendors and data labs, next to Nexdata and Shaip, and like them it publishes no route for outside holders to sell data.
The parent page lists it among dataset vendors and data labs. Related pages:
- Nexdata, another catalog seller
- Shaip, which licenses its own datasets
- Defined.ai, a speech marketplace with Data Partners
- Appen, a listed data services company
- Linguistic Data Consortium, a speech corpus distributor
- ELRA, a language resources distributor
- SpeechData.ai, a speech dataset vendor
- David AI, an audio data vendor
- Company data partnership programs that publish terms
- How to sell a dataset without a vendor
- Alternatives to selling data, including keeping it
- Annotation vendors compared among data labeling companies
More About DataOcean AI
DataOcean AI Speaker Recruitment and Annotator Work
Recording scripts as a speaker or labeling data for DataOcean AI is paid work for the company or its suppliers, not the licensing of company data, and no pay rate or contract for that work is on record.
Such work is compared across companies under AI training jobs and platforms, and whether AI training work is legit is checked platform by platform.
Frequently Asked Questions About DataOcean AI
What is DataOcean AI?
DataOcean AI is the international brand of Beijing Haitian Ruisheng Science Technology Ltd., founded in 2005. Its filings list DATAOCEAN AI subsidiaries in Washington State and Singapore.
Is DataOcean AI the same company as SpeechOcean?
Both names lead to the same group. The parent's Chinese website names dataoceanai.com and keeps speechocean.com addresses in its page source, and its filings list the DATAOCEAN AI subsidiaries.
What is Haitian Ruisheng's stock code?
The stock code is 688787, on the Shanghai Stock Exchange STAR Market. The shares have been listed since 2021-08-13.
Does DataOcean AI buy data from companies?
No program for it is on record. DataOcean AI buys collection and annotation work from suppliers, and no holder terms or payout are on record as of 2026-11-16.
How much do DataOcean AI datasets cost?
No price or customer license for DataOcean AI datasets is on record as of 2026-11-16. Other vendors' stated prices are compared under AI training data prices.
Is DataOcean AI legit?
It is backed by a listed parent that files periodic reports. The half-year report states no major litigation or arbitration for the first half of 2026, and no regulator action is on record as of 2026-11-16.