Archways — agentic threat model

Not certain from the listing — likely utilizes commercial LLMs to parse complex software requirements and match them to vendor capabilities. Primary threats include prompt injection designed to bias software recommendations toward specific vendors or bypass internal compliance checks.

Monitors a massive database of 175,000+ software products and ingests sensitive internal company data (software stack, spend, renewal cycles). Key threats include data poisoning of the vendor database to promote malicious software, and unauthorized exfiltration of the client's proprietary IT stack and financial data.

Not certain from the listing — likely uses a proprietary orchestration framework to coordinate multi-step procurement checks across teams. Threats include insecure tool integration if the agent connects directly to internal IT asset management or financial systems to track spend.

Not certain from the listing — presumably deployed as a multi-tenant SaaS platform. Main threats involve tenant isolation failure, allowing one customer to view another's sensitive software spend, renewal pipelines, and internal compliance gaps.

Not certain from the listing — no details are provided regarding continuous evaluation or guardrails. Gaps here could allow biased recommendations or hallucinated compliance verifications to go unnoticed, leading to the procurement of non-compliant software.

Not certain from the listing — while the agent evaluates compliance for other software, its own security certifications (e.g., SOC2, ISO 27001) are not specified. A lack of robust role-based access control (RBAC) could allow unauthorized internal users to view sensitive financial spend and renewal data.

Not certain from the listing — currently focuses on coordinating human teams. However, future integrations with external vendor-side sales agents or automated procurement marketplaces could introduce trust-abuse threats and cascading negotiation failures.