SECAI Core · phase 14 of 15
AI Vulnerability Database
An open-source knowledge base that collects the data about failures modes for AI models, datasets, and systems
The Explain card
- Plain English
- The AI Vulnerability Database is an open-source knowledge base that collects information about failure modes in AI models, datasets and systems, so practitioners can learn what goes wrong and how.
- Example
- While assessing a vision model, a tester consults the database, finds documented weaknesses affecting that model family, and adds the corresponding evasion tests to the assessment plan.
- Why it matters
- Traditional vulnerability feeds do not cover problems like poisoned datasets or models that fail on certain inputs. A dedicated database fills that gap for AI defenders and red teams.
- Hook
- A CVE list for things that were never just code.
Where it sits in the deck
Phase 14: Incident Response, Evaluation, and Knowledge Bases
When detection fires, teams need structured response plans, scoring frameworks, and curated knowledge bases to triage, measure, and learn from AI security events.
