Unreviewed AI-Suggested Code & Dependencies

Developers or coding agents adopt generated code and suggested packages without review, scanning, or provenance checks.

Impact

Insecure patterns and malicious dependencies enter the codebase through trusted tooling.

Why this severity

Impact High × Likelihood Medium = High

Impact high: vulnerabilities and malware reach production. Likelihood medium: backdoored rules files for coding assistants were demonstrated; most teams still review code.

Impact High: Code execution, credential compromise, exposure of regulated or confidential data, irreversible actions, or system-wide failure.
Likelihood Medium: Publicly demonstrated against real products or in peer-reviewed research, or seen in the wild only with a prior foothold such as stolen credentials.

Method: OWASP Risk Rating Methodology

Evidence

Demonstration: Rules File Backdoor: Supply Chain Attack on AI Coding Assistants (MITRE ATLAS case study AML.CS0041, 2025)
Research: We Have a Package for You! A Comprehensive Analysis of Package Hallucinations by Code Generating LLMs (arXiv 2406.10279, 2024)

Exploited by

Hallucinated Package Squatting Threat · High severity

Mitigated by

Dependency Allowlisting & Package Verification Guardrail · Fixes high risks

Remediation

Require code review for AI-generated changes like any other change.
Run dependency and static analysis on every pull request.
Record which changes were AI-assisted for later audit.

How to verify

PRs adding new dependencies require an approval from a code owner.
SAST and SCA run on all branches, including agent-authored ones.

Framework mapping

OWASP Top 10 for LLM Applications 2025

LLM09:2025 Misinformation
LLM05:2025 Improper Output Handling

OWASP Top 10 for Agentic Applications 2026

ASI05:2026 Unexpected Code Execution

MITRE ATLAS

AML.T0060 Publish Hallucinated Entities

NIST AI RMF 1.0

MANAGE 3.1 Third-party AI risks are monitored and controls applied