aialib: An AI-Aware Threat Library and Generator-Detector Pipeline for LLM-Generated Code
First public release of aialib — a reproducible pipeline and annotated threat library for AI-generated code vulnerabilities. Ships a deterministic generator–detector–annotator pipeline (Bandit, Semgrep, AI-aware risk detector) that turns one YAML prompt manifest into a CSV / SQLite / SARIF dataset. Dataset: 504 annotated findings drawn from 2,144 generated Python samples across four LLM providers (OpenAI gpt-4o-mini, gpt-4.1; Anthropic claude-sonnet-4-20250514; MBZUAI-IFM K2-Think-v2). Each entry carries CWE, OWASP, AI-cause, propagation vector, human–AI factor, and MITRE ATLAS technique labels, plus token and AST fingerprints for deduplicati
Resource information
- Format
- zip
- Original repository
- 20127950
- DOI
- 10.5281/zenodo.20127950
- Credit basis
- Creators in Zenodo record metadata
- Access
- Open files available from the original Zenodo deposit
- Last checked
- Nov 5, 2026
- License
- apache2.0
- Last record update
- Nov 5, 2026
About this resource
First public release of aialib — a reproducible pipeline and annotated threat library for AI-generated code vulnerabilities. Ships a deterministic generator–detector–annotator pipeline (Bandit, Semgrep, AI-aware risk detector) that turns one YAML prompt manifest into a CSV / SQLite / SARIF dataset. Dataset: 504 annotated findings drawn from 2,144 generated Python samples across four LLM providers (OpenAI gpt-4o-mini, gpt-4.1; Anthropic claude-sonnet-4-20250514; MBZUAI-IFM K2-Think-v2). Each entry carries CWE, OWASP, AI-cause, propagation vector, human–AI factor, and MITRE ATLAS technique labels, plus token and AST fingerprints for deduplication. Reproducibility: make run or docker run reproduces the bundled dataset row-for-row under seed 1337; the run manifest records analyzer versions, ruleset hashes, and pipeline version. Licensing: Apache 2.0 for code, CC BY 4.0 for the dataset ( sample_outputs/threat_library.* ). Accompanies the master's thesis "Why AI-Generated Code Is Vulnerable: Cause Taxonomy, Threat Mapping, and CI/CD Security Controls" (MBZUAI, 2026), submitted to the ASE 2026 Tools and Datasets Track.