AI Coding Assistant Supply Chain Risk

Coding assistants and agents pull suggestions, packages, and context from open-source models, public repositories, and MCP servers. An unvetted component can introduce vulnerabilities, license conflicts, or malicious payloads straight into a merge request. PointGuard AI evaluates the AI components engineering teams depend on and documents them in an AI bill of materials, so security teams know what is in the pipeline before it ships.

The AI supply chain behind coding tools changes daily and is rarely reviewed.

• Open-source models and MCP servers are adopted without security review.
• Assistants can suggest hallucinated or typosquatted packages.
• Poisoned models and plugins can carry hidden backdoors.
• License terms on models and datasets can conflict with commercial use.

Supply Chain Security and AI Discovery & Inventory evaluate coding AI components and track their risk over time.

1. Discover coding AI. Identify coding assistants, agents, models, plugins, and MCP servers across developer endpoints, repositories, and CI/CD pipelines.

2. Score each component. Evaluate publisher reputation, known vulnerabilities, unsafe file formats, and license terms. Flag components that fail engineering acceptance criteria.

3. Generate AI-BOMs. Document the models, datasets, libraries, and MCP servers behind each application so security and legal teams can review dependencies.

4. Gate and monitor. Block high-risk components before deployment, re-score when versions change, and alert owners when a dependency's risk rises.

• OWASP Top 10 for LLMs: LLM03 Supply Chain and LLM04 Data and Model Poisoning, covering compromised open-source components and tampered models entering the development pipeline.
• OWASP Top 10 for Agentic Applications: ASI04 Agentic Supply Chain Vulnerabilities and ASI05 Unexpected Code Execution, addressing malicious plugins and MCP servers and code that runs without review.
• NIST AI RMF: GOVERN sets acceptance criteria; MAP identifies dependencies; MEASURE scores component risk; MANAGE blocks unsafe components.