Data protection
We minimise the data used, agree purpose and retention up front, and design for encryption in transit and at rest.
Trust & security
How we protect data, control access, choose where systems run, and keep people responsible for decisions, in every project.
No training on your data
Not without your written agreement, project by project.
Read moreSix controls in every engagement
The details are agreed with your security, legal and operations teams.
Access control
Role-based access, least-privilege service accounts, and integration with your identity provider where available.
Deployment choice
Deploy in your cloud, on private infrastructure, or at the edge for vision workloads, with data location agreed to your policy.
Human oversight
Review queues, approval thresholds, and handover paths keep people responsible for consequential decisions.
Evaluation & monitoring
Acceptance criteria before launch. Quality, drift, latency, and cost monitoring after it.
Auditability
Activity logs, versioned prompts, models, and configurations, and documentation your teams can review.
Your data during a project
Clear rules for each stage, for only as long as the work needs.
- Agree the purpose, the data needed and its owners before anything is shared.
- Work from representative samples with the least personal data possible.
- Keep project data only in environments you approve.
- Give access only to named people, with credentials you can revoke.
- Return or delete project data when the work ends, as agreed.
Run it where it belongs
Your cloud account
Deploy into your AWS, Azure or Google Cloud account under your own identity, network and logging policies.
Private infrastructure
For workloads that must stay inside your network, we plan for private servers and on-premises deployment.
Edge devices
Vision workloads can run next to the cameras, so footage stays on site and only results leave it.
How we build AI responsibly
Transparency
People should know when they are talking to AI. Our voice assistants say so, and every flow offers a route to a person.
Language inclusion
We test quality across the languages, accents and terms your users actually use.
Clear limits
Each system has written limits on what it may answer, access and do. Anything outside them is declined or handed over.
Named owners
Documented decisions, monitoring and a named owner make responsibility clear after launch.
For security and procurement teams
We support vendor reviews with clear documentation.
- Architecture and data-flow diagrams for the proposed solution
- Answers to your security and privacy questionnaires
- Access, logging and retention design for review
- Evaluation plans and acceptance criteria
India’s Digital Personal Data Protection Act, 2023 and sector rules may apply to your use case. We design with your legal and compliance teams. NIM does not give legal advice.
Questions, answered
Something else on your mind? Ask us directly.
Where is our data stored?
In environments agreed with you, often your own cloud account or servers. The location is written into the solution design.
Do you train models on our data?
Not without your explicit, written agreement. Whether project data may be used to fine-tune a model for your solution is decided per engagement.
Can we see what the AI did and why?
Yes. We plan activity logs and configuration versioning so your teams can review actions, inputs and outcomes, within the retention rules you set.
Who can access our systems during the project?
Only named team members, using access you grant and can revoke. We prefer least-privilege, project-specific credentials.
