Ascendo AI Resolution AI Agents — agentic threat model

Not certain from the listing — The specific LLMs or ML models used are not disclosed. General threat: Adversarial prompt injection via customer-facing chat channels could manipulate the underlying model to bypass safety guardrails or trigger unauthorized RPA actions.

Not certain from the listing — The exact vector database or RAG architecture is not specified. General threat: Poisoning of the knowledge base or systems of record could lead to incorrect spare parts planning or malicious customer support instructions.

Not certain from the listing — The orchestration framework is not named. However, the use of Cognitive RPA and integration with systems of record poses a high risk of tool misuse or insecure tool execution if prompt injection occurs.

Not certain from the listing — Hosting environment (cloud/on-prem) is not detailed. General threat: Compromise of the API integrations or RPA runners could allow lateral movement into connected enterprise systems of record.

Not certain from the listing — No specific guardrails or evaluation frameworks are mentioned. General threat: Lack of real-time monitoring for anomalous RPA actions could lead to undetected automated transaction failures.

Not certain from the listing — No compliance certifications (e.g., SOC 2, ISO 27001) or specific RBAC mechanisms are detailed. General threat: Insufficient access controls between the agent and connected ERP/CRM systems could violate least privilege.

Not certain from the listing — While 'Agents' is plural, explicit multi-agent coordination protocols are not described. General threat: Cascading failures if a compromised customer-facing agent interacts maliciously with back-end planning agents.