Make — agentic threat model

Not certain from the listing — The description focuses on visual workflow building and app integrations but does not specify the underlying foundation models (LLMs) used, leaving potential risks like prompt injection, model misalignment, or adversarial manipulation unaddressed at this layer.

Not certain from the listing — While real-time data synchronization and data transfer limits are mentioned, there is no explicit detail on vector stores, RAG pipelines, or data lineage, making data poisoning or embedding inversion risks difficult to assess directly.

Make's core framework relies on visual workflow building and decision-based logic to orchestrate integrations. The primary threats here are insecure tool integration and tool misuse, where compromised or poorly configured logic blocks can trigger unintended actions across connected APIs.

Not certain from the listing — The platform is closed-source and cloud-hosted, but the listing does not provide details on execution sandboxing, secrets management for connected app credentials, or network isolation controls.

Not certain from the listing — The listing does not detail specific AI-focused evaluation, guardrails, or real-time anomaly detection mechanisms, though standard execution logging is typical for workflow engines.

The listing explicitly highlights 'Customizable roles and permissions' and 'Shared workspaces for team collaboration', indicating built-in access control mechanisms to manage user authorization and mitigate unauthorized workflow modifications.

Make features an 'Extensive template library' and a 'Wide range of app integrations', creating an ecosystem where third-party templates or malicious app connectors could introduce supply chain risks or cascading failures across integrated services.