HATCH CANVAS — agentic threat model
Hatch Canvas presents a moderate-to-high risk profile due to its ability to generate, execute, and publish interactive web applications and code directly from an infinite canvas. The primary threats stem from potential prompt injection leading to malicious code generation (XSS) and the lack of visible sandboxing or content moderation controls.
OWASP AIVSS score rationale
| Autonomy of Action | 0.40 | |
| Goal-Driven Planning | 0.40 | |
| Self-Modification | 0.10 | |
| Dynamic Tool Use | 0.60 | |
| Persistent Memory | 0.50 | |
| Contextual Awareness | 0.80 | |
| Dynamic Identity | 0.10 | |
| Multi-Agent Interactions | 0.50 | |
| Non-Determinism | 0.70 | |
| Opacity & Reflexivity | 0.60 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Not certain from the listing — The specific LLMs or multi-modal models powering Hatch Canvas are not disclosed. Threats include model reprogramming or prompt injection leading to malicious code generation or inappropriate image synthesis.
Not certain from the listing — The mechanism for storing canvas state, user interactions, and RAG data is unspecified. Risks include data exfiltration of private canvas designs or poisoning of the context-aware memory.
Hatch Canvas orchestrates multiple AI tools and generates interactive components/code. Insecure tool integration or prompt injection could lead to the generation of malicious scripts (XSS) executed directly within the collaborative canvas.
Not certain from the listing — The hosting environment for running generated web apps, games, and the canvas itself is not detailed. If sandboxing is weak, generated apps could lead to container escape or lateral movement.
Not certain from the listing — No guardrails, content moderation, or logging mechanisms are mentioned. This creates blind spots regarding the generation of abusive, copyrighted, or malicious multi-modal content.
Not certain from the listing — Access controls for private, public, or remixable canvases are not detailed. Lack of robust authentication and authorization could allow unauthorized modification or exfiltration of proprietary designs.
The platform supports multi-agent collaboration ('all AI tools work together') and remixable public sharing. This introduces risks of cascading failures, where a compromised or malicious remixed canvas infects other collaborating agents or users.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).