DPDP Act AI training data compliance, built into every clip
Every frame is captured with consent. Every clip has a paper trail. Collection is consent-first and approval-gated, with workflows designed around India's Digital Personal Data Protection Act, 2023.
Real-world training data is only useful if your legal team can sign it off. Footage scraped from the web or recorded without clear consent creates risk that no model improvement is worth. For robotics AI labs buying data from another country, the questions are always the same: who is in this footage, did they agree, where was it recorded, and can you prove it?
Our answer is a process where consent and traceability come first. Every participant is briefed in their own language before a single frame is recorded, every site is covered by a written agreement, and every clip carries the metadata needed to trace it back to who captured it, where, and under which consent. That is DPDP Act AI training data compliance built into the workflow, not added afterwards.
What every project includes
- Written informed consent in Bengali, Hindi or English
- Site agreements with owners/managers
- DPDP Act-aware workflows: collection, storage and access designed with India's Digital Personal Data Protection Act, 2023 in mind
- Face-blurring on request, before delivery, when your project or a participant needs it
- Used only for your model training. Footage is never resold.
1. Informed consent, in the participant's language
Before recording starts, each participant is briefed on what will be recorded, why, how the footage will be used and how it is protected. Consent is written and given in Bengali, Hindi or English, whichever the participant is most comfortable with. Nobody is recorded before their consent is signed.
2. Written site agreements
Every factory, warehouse, kitchen or workshop where we record is covered by a written agreement with its owner or manager. The agreement sets out what will be recorded, when and where, so there is no doubt about the right to capture footage on the premises. Site onboarding and access negotiation are part of our egocentric data collection service.
3. An audit trail on every clip
Each clip carries the metadata needed to trace it: operator ID, site, task and consent reference. If your legal, privacy or data team needs to answer "who is in this clip and did they agree?", the answer is in the record. The same approval-gate discipline we use in SteelBox applies: AI proposes, humans confirm, every decision is traceable.
4. DPDP Act-aware workflows
India's Digital Personal Data Protection Act, 2023 governs how personal data, including video of identifiable people, is collected and processed. Our collection, storage and access workflows are designed with the Act in mind: consent before collection, clear purpose, restricted access and secure transfer. How we handle personal data on this website is set out in our Privacy Policy.
5. Face-blurring on request
Some projects need identifiable faces removed, and some participants ask for it. Face-blurring is available on request and is applied before delivery.
6. Used only for your model training
Footage is collected for your model training under a signed agreement and is never resold or reused for other clients. Every participant has given written consent, every site has a written agreement, and every clip carries the metadata and audit trail your legal team needs to sign off.
What this means for your team
- Legal gets consent and site records that map to every clip.
- Data gets consistent metadata that can be filtered by operator, site and task.
- Research gets footage it can actually use, from environments no one else has captured.
Compliance across every service
The same process applies whether you buy VLA training data, imitation learning data or humanoid robot data. Need consented data your legal team can sign off? Start with the 50–100 hour pilot.