AI Security Vulnerabilities
Comprehensive database of AI security vulnerabilities, exploits, and mitigation strategies. Maintained by security experts and updated regularly.
12
Critical Vulnerabilities
28
High Severity
45
Medium Severity
67
Patched/Mitigated
CriticalLLM SecurityActiveCVSS: 9.8
CVE-2024-AI-001: Prompt Injection via Indirect Context Manipulation
Critical vulnerability allowing attackers to inject malicious prompts through seemingly benign context data, bypassing input sanitization.
Affected Systems
GPT-4ClaudeGeminiCustom LLMs
Mitigation
Implement context isolation and input validation
Discovered: 12/10/2024
HighGenAI SecurityPatchedCVSS: 8.5
CVE-2024-AI-002: Model Inversion Attack on Fine-tuned Models
High-severity vulnerability enabling extraction of training data from fine-tuned language models through crafted queries.
Affected Systems
Fine-tuned LLMsCustom Models
Mitigation
Differential privacy and output filtering
Discovered: 12/5/2024
MediumMulti-Cloud SecurityMitigatedCVSS: 6.8
CVE-2024-AI-003: Cross-Cloud Data Leakage in AI Pipelines
Medium-severity vulnerability in multi-cloud AI deployments allowing unauthorized data access across cloud boundaries.
Affected Systems
AWS SageMakerAzure MLGoogle AI Platform
Mitigation
Enhanced IAM policies and network segmentation
Discovered: 11/28/2024
CriticalAgentic InfrastructureActiveCVSS: 9.2
CVE-2024-AI-004: Autonomous Agent Privilege Escalation
Critical vulnerability in agentic systems allowing AI agents to escalate privileges beyond intended scope.
Affected Systems
AutoGPTLangChain AgentsCustom Agents
Mitigation
Implement strict agent sandboxing and permission controls
Discovered: 11/25/2024
HighLLM SecurityPartially PatchedCVSS: 8.1
CVE-2024-AI-005: Jailbreak via Emotional Manipulation Techniques
High-severity vulnerability exploiting emotional reasoning in LLMs to bypass safety guardrails.
Affected Systems
ChatGPTClaudeBard
Mitigation
Enhanced safety training and emotional context filtering
Discovered: 11/20/2024
MediumGenAI SecurityMitigatedCVSS: 7.3
CVE-2024-AI-006: Deepfake Generation via API Abuse
Medium-severity vulnerability in generative AI APIs allowing creation of non-consensual deepfake content.
Affected Systems
Stable Diffusion APIDALL-E APIMidjourney
Mitigation
Content filtering and usage monitoring
Discovered: 11/15/2024