Corporate LLM/RAG with Kaspersky AIST security audit
Not a chatbot — secure corporate AI: data leakage control, prompt injection, RAG risks, usage policy and exec report. Bober AI Systems' AI Security Gateway + Kaspersky AIST services.
What the client buys
- — Control of data leakage via the model
- — Prompt injection and jailbreak tests
- — RAG vulnerability and knowledge-base poisoning analysis
- — Check that the agent cannot exceed permissions
- — Safe AI usage policy
- — Risk report for leadership
Bober AI Systems
- — Private LLM or RAG deployment
- — Bitrix24, 1C, mail and document integration
- — Agent access control and permissions
- — Logging of AI actions
- — Tool and API hardening
- — System prep for Kaspersky AIST assessment
Kaspersky
- — Threat model for the AI solution (AIST)
- — Model and data analysis (input / training)
- — Resilience tests and AI security monitoring
- — AI security and regulation consulting
- — Report and recommendations
Architecture
- 01 User → LLM/RAG in a closed contour
- 02 Integrations: Bitrix24 / 1C / documents / APIs
- 03 Audit log + RBAC + agent policies
- 04 Artifact pack → Kaspersky AIST assessment
- 05 Risk report → leadership / security
Most direct Security for AI. Confirm with Kaspersky partner manager whether Registered Partners can resell/register AIST projects.