Controlled environment
Infrastructure and access are defined for the engagement.
Private AI & data security
Use AI without surrendering control of the information.
Public AI services are not the only way to analyse large volumes of sensitive material. The 10th Man can work with locally deployed models and controlled environments selected for the engagement.
The principle
Your information should stay inside boundaries you understand.
Local AI can reduce exposure to public services, but local does not automatically mean secure. The environment, access, storage, data handling and operating process all matter.
The right model is selected after understanding the material, the task, the sensitivity and the client's own controls.
Source traceability
Material findings can be checked against the documents that support them.
Human accountability
AI assists the work. It does not own the conclusion.
Engagement models
Different controls for different work.
The appropriate design depends on sensitivity, volume, location and the client's technical environment.
The 10th Man local infrastructure
Analysis on controlled local systems, with defined storage, access and deletion arrangements.
Client-hosted environment
Models and document analysis operated inside infrastructure controlled by the client.
Client-site deployment
Work performed within the client's physical and technical environment where required.
Air-gapped environment
Offline analysis for highly sensitive engagements where internet-connected processing is inappropriate.
Controls that may form part of the design
Security is a system, not a label.
- Encrypted storage and controlled access
- Local document analysis
- Defined retention and deletion arrangements
- Source-level verification of material findings
- No unnecessary transfer to public AI services
- Human review before conclusions are delivered
Important
No system is risk-free. The 10th Man does not claim that local AI is automatically secure, “unhackable” or appropriate for every engagement. Controls, limitations and responsibilities should be agreed before sensitive material is provided.
Sensitive, high-volume evidence?
Discuss the boundary before sharing the material.
An initial conversation can establish the task and the right handling model.