Data boundaries
Data classification, approved uses, prohibited data, regions, retention, deletion, and client versus consultancy environments.
Security requirements are part of the system design.
AI projects can cross sensitive data, external model providers, production systems, and specialized subcontractors. Those boundaries are defined before access and reflected in the architecture and engagement terms.
Current position
Specific controls. Precise claims.
Tensor State Group does not present unearned certifications or blanket compliance claims. Security and procurement responses are scoped to the actual entity, engagement, vendors, data, and client environment.
Access and delivery
Named access, least privilege, MFA, repository and environment ownership, subcontractor access, and offboarding.
AI-system risk
Provider approval, model and dataset provenance, prompt/data leakage, evaluation, misuse, failure handling, and human oversight.
Production operation
Logging, observability, change management, incident escalation, backups, recovery, and post-handover ownership.
During contracting
Procurement topics are addressed before delivery begins.
Commercial
- NDA, MSA, and statement of work
- Roles, dependencies, acceptance, and change control
- IP, pre-existing materials, and licensing
- Insurance and vendor documentation where applicable
Data and security
- Data-processing and security requirements
- Approved infrastructure, models, and subprocessors
- Incident, retention, return, and deletion terms
- Client questionnaire and review process
Delivery assurance
- Governance and technical reporting
- Evaluation and production-readiness criteria
- Documentation and knowledge transfer
- Operating owner and support boundaries
Do not send credentials, confidential datasets, controlled material, or detailed vulnerability information through the public scheduling flow. Request an approved communication channel during initial contact.