AI Security Reference
PromptRiskDB
PromptRiskDB turns public AI security and vulnerability datasets into linked pages about attacks, risks, mitigations, case studies, and CVE context.
Recent Records
Newly indexed AI security records from public datasets.
AI Security Techniques
Ways AI systems can be attacked or misused, with examples and defenses.
AI attack method AML.T0112.001; evidence level: feasible; 5 related AI risks
AI attack method AML.T0034.001; evidence level: feasible; 4 related AI risks
AI attack method AML.T0034.000; evidence level: feasible
AI Risks
Common risks that can happen when AI systems are built, deployed, or used.
AI risk record; domain: 2. Privacy & Security; subdomain: 2.0 > Privacy & Security; when it can happen: 2 - Post-deployment
AI risk record; domain: 7. AI System Safety, Failures, & Limitations; subdomain: 7.3 > Lack of capability or robustness; when it can happen: 2 - Post-deployment
AI risk record; domain: 4. Malicious Actors & Misuse; subdomain: 4.3 > Fraud, scams, and targeted manipulation; when it can happen: 2 - Post-deployment
AI Mitigations
Practical safeguards that help reduce AI security and safety risks.
7 AI defenses grouped under the Policy category.
10 AI defenses grouped under the Monitoring and Maintenance ML lifecycle stage.
10 AI defenses grouped under the ML Model Evaluation ML lifecycle stage.
AI Case Studies
Real and documented examples of AI-related security incidents.
AI security case study AML.CS0042; type: incident; date: 2025-07-01; target: OpenAI Assistants API; actor: Unknown Threat Actor; 1 case steps
AI security case study AML.CS0015; type: incident; date: 2022-12-25; target: PyTorch; actor: Unknown; 3 case steps
AI security case study AML.CS0007; type: exercise; date: 2019-08-22; target: OpenAI GPT-2; actor: Researchers at Brown University; 5 case steps
AI Vulnerabilities
Software vulnerabilities connected to AI systems, incidents, or known exploitation.
CVE-2026-25253 vulnerability profile; not listed by CISA as known exploited; 0 related AI attack methods; 1 related case studies
CVE-2025-8217 vulnerability profile; not listed by CISA as known exploited; 0 related AI attack methods; 1 related case studies
CVE-2023-6021 vulnerability profile; not listed by CISA as known exploited; 0 related AI attack methods; 1 related case studies
ATLAS Tactics
The main goals attackers may pursue when targeting AI systems.
MITRE ATT&CK AI Context
How AI security issues connect to broader cybersecurity attack techniques.