Privacy Policy
Version 0.6, effective 21 September 2026. This Policy is written to the Information Technology Act, 2000 and the Information Technology (Reasonable Security Practices and Procedures and Sensitive Personal Data or Information) Rules, 2011 as they apply today, and to the Digital Personal Data Protection Act, 2023 and the rules under it as its provisions commence.
This Policy explains what personal data Tensordata Global Private Limited ("TakeOK", "we", "us") collects when you use the TakeOK console, websites and APIs (the "Service"), why, who it is shared with, how long it is kept, and the rights you have. It is a notice, not a contract term, and accepting the Terms of Use is not consent to processing; where consent is the basis for a processing activity we ask for it separately, at the point of collection, for the purpose stated there. Capitalised terms have the meaning given in the Terms of Use.
1. Who is responsible for what
1.1 Roles depend on the processing activity, not on who uploaded the data:
| Activity | Who determines the purpose | Our role |
|---|---|---|
| Your account, sign-in, approval, workspace and team membership; billing, invoices and tax records; running the Service and diagnosing faults | We do | Data fiduciary; the individual whose data it is — you, or the person at your organisation who holds the account — is the data principal |
| Safety screening, abuse and fraud investigation, grievance handling, legal compliance | We do, for our own purposes, even where the content concerned is customer-controlled | Data fiduciary for that processing |
| Content supplied for a project: faces, voices, likenesses, clips, scripts | Whoever decided to make the project: you, your employer or organisation, or your client. A person uploading their own face for their employer's project is usually acting within the employer's arrangement, and the employer determines the purpose | Data processor for that purpose, under the Data Processing Agreement at https://takeok.tensordata.com/legal/dpa (the "DPA"); where you act as processor for your client, we are your sub-processor |
| Content you make for yourself, on your own account, for your own purpose | You decide what to make and why; we decide how the Service processes it |
Roles follow who determines the purpose and means of a processing activity. The DPA documents the processor or sub-processor relationship where one exists; it does not create or override it.
1.2 Where you supply other people's personal data, you or your client decide why it is processed; the DPA sets out our instructions, the sub-processors we use, security, incident notice, deletion, and how we help you answer requests from those people. The Terms of Use, Section 3.4, set out what you promise us about lawful basis and consent for that data.
1.3 Contact. Tensordata Global Private Limited, Bengaluru, Karnataka, India. Privacy: privacy@tensordata.com. Grievance Officer: Grievance Officer, grievance@tensordata.com, [phone not set], Bengaluru, Karnataka, India. We acknowledge a grievance within 24 hours and resolve it within 7 days, with the shorter statutory timelines in Section 5 of the Terms of Use for particular content. If you are dissatisfied you may appeal to the Grievance Appellate Committee established under the Information Technology Rules, and, once the relevant provisions of the Digital Personal Data Protection Act commence, complain to the Data Protection Board of India.
2. What we collect
2.1 Account data. Name, email address, password (stored as a one-way hash), organisation, role, the approval status of your account, and the workspaces, teams and projects you belong to.
2.2 Content data. Everything you supply to make media: prompts, scripts and briefs; reference images, including photographs of faces and products; audio recordings, including voice samples used to clone a voice; video clips you upload for analysis, editing or as motion references; brand assets; settings and choices. Also what the Service derives and produces: generated media, thumbnails and extracted frames, contact sheets, transcripts and word timings, descriptions and analyses produced from your uploads, cloned-voice clips and any voice profile or embedding a speech model derives, and the parameters recorded with each render so it can be traced and reproduced.
2.3 Sensitive content. Facial images and voice recordings can identify a person and are treated by us as sensitive. We process them only to: perform the operation you request, such as keeping a face consistent across shots or speaking a line in a chosen voice; review a complaint or a suspected violation of the Terms; provide support you have asked for, with your permission to look at the content; and preserve evidence where the law or a legal order requires. We do not use them to identify people across accounts, to build a database of faces or voices, or to train models.
2.4 Usage and technical data. Server logs including IP address, browser and device information, pages and features used, timestamps, render durations and outcomes, error reports, and the identifiers of the jobs our platform runs.
2.5 Payment data. Where the Service is paid: plan, credits, invoices, GSTIN and billing address. The Service is not yet paid. When payments begin, card and bank details will be collected and stored by a payment gateway, not by us, and this Policy will name it and what we receive from it before any payment is taken.
2.6 Communications. Messages you send to support or the Grievance Officer, consent documents you supply under the Terms, and our replies.
2.7 Cookies. Cookies for sign-in, session integrity, security and your saved preferences, set because the Service you asked for cannot work without them; they are processed on the same ground as providing the Service in Section 3.1. We do not use analytics cookies; where we do, they are listed in the cookie notice, are off until you accept them, and can be refused without affecting the Service.
We do not knowingly collect data from anyone under 18.
3. Why we process it, and the legal ground
3.1 Ground today and from commencement. Until the substantive provisions of the Digital Personal Data Protection Act, 2023 commence for us — the notified schedule brings most duties and rights into force in May 2027 — the Information Technology Act, 2000 and the 2011 Rules govern, and we process personal data with your consent for the purposes stated at collection and as the law otherwise permits. From commencement, each activity rests on consent given for a specified purpose, or on one of the "certain legitimate uses" the Act lists. The table says which:
| Purpose | Ground under the Digital Personal Data Protection Act, 2023 | Until commencement |
|---|---|---|
| Providing the Service you signed up for: rendering, editing, storing and delivering media; keeping faces, products and voices consistent across a project; transcription and translation; your history | Your consent, given at sign-up for these specified purposes and confirmed where the Service asks at upload (Section 6) | Consent under the 2011 Rules and the contract with you |
| Processing other people's personal data that you or your client supply | Not our ground to hold: the data fiduciary — you or your client — holds it, and we process on written instructions as processor under Section 8(2) and the DPA | Consent obtained by you; the 2011 Rules |
| Keeping the Service safe: automated screening of prompts, uploads and outputs; detecting abuse, fraud and security incidents; investigating complaints | For your own data: your consent at sign-up to safety screening as a specified purpose of the Service. For the data of a person depicted in content a customer supplies: that person's consent, obtained by the customer through the release the Terms require, which covers our safety review and grievance handling; and, for investigations where consent is absent, whatever statutory ground or exemption applies. The security practices, due-diligence and record-keeping duties the Information Technology Act and Rules and the CERT-In directions impose are obligations on us that shape this processing; they are not themselves a Section 7 ground | Due-diligence duties under the Information Technology Rules; the 2011 Rules |
| Disclosing personal data to the State where a law obliges us to (Section 7(d)); complying with a judgment, decree or order of a court or authority (Section 7(e)) | The legitimate uses in Sections 7(d) and 7(e), each for the disclosure or compliance the law or order actually requires and nothing wider | Legal obligation |
| Billing and accounts: payments, tax invoices, fraud prevention | Your consent at purchase for these specified purposes. Invoices and payment records are then kept as restricted copies under the Act's erasure exception for retention required by law (Section 5) | Consent and legal obligation |
| Running and improving the Service: measuring render times, diagnosing errors, capacity planning | Your consent at sign-up for these specified purposes. Aggregated statistics from which no person or content can be identified are not personal data | Consent |
| Communicating with you about the Service: approvals, render completion, invoices, security notices, changes to terms | Your consent at sign-up; a notice the law itself requires is given under that law | Consent |
| Marketing email | Separate, optional consent, withdrawable at any time | Separate consent |
| Reporting security incidents to CERT-In, and, from commencement, personal-data breaches to the Data Protection Board | Section 7(d), for the report the direction or the Act requires | Legal obligation under the CERT-In directions |
Where the table says "your consent", the consent is asked for separately from acceptance of the Terms, names the purpose, and is recorded with its version and time. Consent to a purpose covers every upload made for that purpose; a confirmation at upload records that the specific file is supplied for it.
3.2 We do not use your content to train, fine-tune or evaluate machine-learning models. Our generation, speech and vision models run on our own platform; no training, fine-tuning, evaluation or dataset-export pipeline exists on it, and there is no setting by which one could be turned on.
3.3 Withdrawing consent. You may withdraw consent at any time through the account settings or by writing to privacy@tensordata.com. Withdrawal is as easy as giving consent and does not affect processing already done. Withdrawing consent is not the same as cancelling a job; both are available, and withdrawal has these effects: operations still queued for that purpose are cancelled and not charged; an operation already rendering is stopped as far as technically practicable, and compute already consumed is charged as Section 6.3 of the Terms of Use provides, never above the lower of the estimate and the operation's ceiling; a completed operation's charge stands, subject to any refund the law requires; stored media that exists only for that purpose is removed from active use at once and deleted within the periods in Section 5; features that need the data stop working; and records we must keep by law, listed in Section 5, are kept as restricted copies for the stated period only.
4. Who we share it with
4.1 Where your content is processed, and the vendors involved. Prompts, reference images, audio and video are processed by generation, analysis, transcription, speech and moderation models that run on our own platform in Bengaluru, operated by us on our own hardware and network; no third party operates a model for us and no model licensor receives your data. Our platform processes your content only to perform the operation, return and store its result for the retention stated in the register, and meter usage. The vendors that do receive data, and what they receive, are listed in the register maintained at https://takeok.tensordata.com/legal/privacy and reproduced here:
| Provider (legal entity) | Function | Data it receives | Processing location | Retention on the provider's side | Training on your content |
|---|---|---|---|---|---|
| Operated by us, on our own platform | Video and image generation, stored-clip upscaling | Prompts, reference images and clips, audio references, generated media | Bengaluru, India | Inputs 30 days after upload, outputs 90 days after completion; job records including prompt text currently without a purge | No — no training pipeline exists on our platform |
| Operated by us, on our own platform | Voice cloning and speech synthesis | Voice samples, spoken lines, transcripts | Bengaluru, India | Voice sample 30 days as an input; no voice profile or embedding is ever stored; the transcript with the job record | No |
| Operated by us, on our own platform | Prompt drafting, content screening, image reading, style analysis, clip analysis | Prompts, uploaded images and frames, transcripts | Bengaluru, India | Nothing retained: request logging off, no database write, context discarded when the request ends | No |
| Operated by us | Transcription and word timing | Audio extracted from clips | Our own servers | Not retained after the reply | No |
| Operated by us, on our own platform | Object storage and the managed database behind the Service | All content data, account and billing data | Bengaluru, India | Until we delete it; media is in no offsite backup; database backups per Section 5 | No |
| Operated by us (our own mail servers) | Transactional email | Email address, subject, message content, delivery events | Bengaluru, India | 30 days | No |
| Backblaze, Inc. (our backup vendor) | Offsite backup of our platform's control-plane database: job records including prompt text and voice-clone transcripts, never media files | Encrypted database dumps | Netherlands, being moved to our own disaster-recovery site in India | 30 days, encrypted with AES-256 before leaving India, key held in India | No |
No payment gateway is in use; one will be added here before any payment is taken. No external error-monitoring or analytics vendor is used; errors are logged on our own systems in India. Our staff-alerting and support-mailbox vendors (Telegram and Microsoft 365) carry organisation and resource identifiers only, never your content.
Where a model is operated by us or by an affiliate rather than a vendor, the register says so, and the licensor of the model receives no data.
4.2 Cross-border processing. All generation, analysis, speech and storage take place in Bengaluru, India. One path leaves India today: our encrypted backup of our platform's control-plane database — job records including prompt text and any voice-clone transcript, never media files, voice samples or outputs — to our backup vendor in the Netherlands, as the register states. We are moving that backup to a disaster-recovery site of our own in India; when the move is complete this section and the register will say that all data stays in India. Beyond that, providers may process data outside India only as the register states. Today, under Rule 7 of the 2011 Rules, sensitive personal data leaves India only with your consent or where the transfer is necessary to perform the contract with you, and only to a recipient bound to the same level of protection; each provider in the register is so bound. From commencement of the Digital Personal Data Protection Act, 2023, transfers are also subject to any restriction the Central Government notifies under it.
4.3 Members of your spaces and administrators. Content you place in a shared workspace, team or project is visible to its members. Administrators of your organisation's account can see account details and usage. Anyone holding a public sharing link you create can view what it points to.
4.4 Authorities and legal process. Where a court order, a lawful request from a government agency, or the law requires (Sections 7(d) and 7(e)). Beyond that, we disclose personal data to prevent serious harm, to protect a person's safety, or to establish, exercise or defend a legal claim only where the law provides a ground or an exemption for that disclosure, or with the consent of the person concerned. A belief that disclosure would help enforce the Terms is not, on its own, a ground.
4.5 Business transfers. If we merge, are acquired, or sell assets, personal data may transfer to the successor under this Policy, and we will tell you.
We do not sell personal data, and we do not share content with advertisers.
5. How long we keep it
| Data | Retention |
|---|---|
| Account data while the account is open | Life of the account |
| Registration information after closure | 180 days, as the Information Technology Rules require of intermediaries, then deleted |
| Input and Output, including thumbnails, extracted frames, contact sheets, transcripts, word timings and analyses | While your account is open; items you move to the Trash are deleted after 30 days; on closure, deleted within 30 days |
| Voice samples, cloned-voice clips and any voice profile or embedding | As Input. On your request they are removed from active use at once and deleted from our systems within 7 days; the platform's copy of the sample is deleted 30 days after upload, and we keep no voice profile, embedding or other derived representation of a voice at any time; backup copies expire per the Backups row |
| Job identifiers and render parameters | 12 months after the render, for tracing and support. Our platform's own job records, including prompt text and any voice-clone transcript, are kept without a purge today; a purge on the same 30-day clock as inputs is being built, and this row will say when it is in place |
| Server, security and access logs | 180 days, kept within India, as the CERT-In directions of April 2022 require; one year for the traffic and processing records the Digital Personal Data Protection Rules cover once they commence; longer where an incident is under investigation |
| Payment records and invoices | 8 years, as tax and company law require |
| Removed content and the records of a complaint, takedown or legal order | 180 days after removal, or longer where a court or authorised agency requires |
| Consent documents you supply under the Terms | Life of the related content plus 180 days |
| Backups | Rolling, expiring within 90 days; data restored from a backup after its deletion is deleted again within 7 days |
Two kinds of keeping. "Deleted" in this table means removed from active use and erased from our systems within the stated period. Where the law requires us to keep a record for longer — the 180-day intermediary periods above; tax and company law; a legal hold while a claim, complaint or investigation is pending; and, from commencement of the Digital Personal Data Protection Rules, the one-year retention that Rules 6(1)(e) and 8(3) require of the personal data, traffic data and processing records they cover — the data is kept as a restricted copy, inaccessible to ordinary operations, used only for that purpose, and erased when the period ends. Deletion in the console is immediate for the console's own copies. The platform beneath it deletes media inputs 30 days after upload and outputs 90 days after completion; the console's deletion does not yet reach those copies sooner, and the platform's job records including prompt text are kept until the purge being built is in place. The register states these periods.
"De-identified" statistics in this Policy mean aggregated counts and durations from which no content, face, voice or account identifier can be recovered; a face or a voice is personal data whether or not an account identifier is attached to it, and we never treat it as anonymous.
6. How we protect it
Data is encrypted in transit on every connection you reach; traffic inside our private cluster network is not separately encrypted. Credentials and other secrets are encrypted at rest. Media, prompts and database records are not encrypted at rest on our platform storage today; our offsite database backups are encrypted before they leave India. We say so rather than claim otherwise, and this section will be updated when at-rest encryption is in place. Access is limited to staff and contractors who need it, under confidentiality obligations, and is logged. Content is private to your account and the spaces you share it in, as Section 4.3 describes. We test and update our safeguards under the reasonable security practices the 2011 Rules require.
If a cyber-security incident of a kind the CERT-In directions cover affects the Service, we report it to CERT-In within 6 hours of noticing it. If a personal-data breach affects you, we notify you and, from commencement of the relevant provisions, the Data Protection Board, in the form and time the law requires.
7. Your rights
7.1 Now. We offer you, voluntarily and in line with the 2011 Rules: access to the personal data we hold about you; correction of anything inaccurate; deletion of your account and content as Section 5 describes; withdrawal of consent under Section 3.3; and the grievance route in Section 1.3.
7.2 From commencement of the Digital Personal Data Protection Act, 2023. You have the rights the Act gives a data principal: to a summary of the personal data we process and the fiduciaries and processors it is shared with; to correction, completion, updating and erasure; to grievance redressal; to nominate a person to exercise your rights if you die or are incapacitated; and to withdraw consent. You have corresponding duties under the Act, including not to impersonate another person or to supply false particulars.
7.3 How to exercise a right. Use the account settings where the Service offers them, or write to privacy@tensordata.com from the email on your account. We may ask you to verify your identity. We respond within 30 days, or sooner where the law requires, and tell you if we cannot act on a request and why.
7.4 Other people's content. If your face, voice or other personal data appears in content that another user uploaded, write to the Grievance Officer. We act on the complaint under Section 5 of the Terms of Use and may disable the content while it is examined. We tell the uploader what was disabled and why, except where doing so could endanger you or another person, breach a confidentiality obligation or a legal order, or compromise an investigation.
7.5 Users outside India. The Service is offered in India. If we make it available to users in the European Economic Area or the United Kingdom, the additional rights and disclosures the GDPR and the UK GDPR require, including restriction, objection, portability, the lawful bases we rely on there, the transfer mechanism we use and the supervisory authority you may complain to, are set out in Annex A [to be drafted only if those markets are in scope; counsel to advise on applicability, which depends on targeting and establishment, not on a user's location alone].
8. Children
The Service is for adults, and the Terms prohibit uploading the face, voice, likeness or other personal data of any real person under 18 — there is no incidental-appearance exception, so we do not operate a parental-consent process. We do not knowingly process a child's personal data; if you believe a child's data has been processed, write to privacy@tensordata.com and we will remove it from active use at once and delete it under Section 5. We do not track children or target advertising at them. Should we ever permit a controlled exception, this Policy and the Terms will say so together and the verifiable parental-consent process the Digital Personal Data Protection Rules require will be in place first.
9. Automated processing
Content is screened by automated systems before and after rendering to enforce the Terms and the law. An automated refusal can be appealed to the Grievance Officer, who is a person. We do not make decisions with legal or similarly significant effects about you by automated means alone.
10. Changes to this Policy
We may update this Policy. Material changes are notified by email or in the Service at least 15 days before they take effect, and where a change needs fresh consent we ask for it. The date at the top shows the current version; earlier versions are available on request.
11. Contact
Tensordata Global Private Limited, Bengaluru, Karnataka, India. Privacy: privacy@tensordata.com. Grievance Officer: Grievance Officer, grievance@tensordata.com, [phone not set].