Mogoj AI — agentic threat model

Not certain from the listing — The description mentions conversational agents and RAG but does not specify the underlying foundation models used by Chetan or Kriya, leaving threats like model-specific backdoors or adversarial vulnerabilities unquantified.

Gyankosh serves as a unified knowledge base with RAG and vector store capabilities handling documents, media, files, spreadsheets, and presentations. This central repository is highly vulnerable to knowledge-base poisoning, unauthorized data retrieval, and embedding inversion attacks.

Chetan provides a composable framework for building scalable agents, while Kriya offers an SDK and runtime for real-world interactions. Vulnerabilities here include insecure tool integration, framework-level prompt injection, and unauthorized execution of bundled functions.

Not certain from the listing — The deployment model, hosting environments, and sandboxing capabilities of the Kriya runtime are not detailed, making it difficult to assess container escape or privilege escalation risks.

Not certain from the listing — There is no mention of built-in evaluation, monitoring, logging, or guardrail mechanisms to detect drift, anomalous agent behavior, or malicious inputs within the ecosystem.

Not certain from the listing — The listing does not outline specific identity management, authorization policies, or compliance alignments (such as SOC2 or ISO) for the workspace or SDK.

Mitra facilitates collaboration between teams and multiple agents. This multi-agent workspace introduces significant risks of agent-to-agent trust abuse, cascading failures, and horizontal privilege escalation if one agent in the workspace is compromised.