LIVE · THREAT INTELLIGENCE CENTER

AI Security Threat Monitor

Real-time surveillance of AI exploits, prompt injection attacks, supply-chain compromises, and vulnerabilities in vibe-coded apps — pulled from 9 live intelligence sources and explained in plain language so any builder knows what to do.

SourcesSecurity SignalsHacker NewsReddit SecurityCISA KEVNVD CVE 2.0OSV.devGitHub AdvisoriesSecurity RSSVBS Registry
0 items
MITRE ATLAS — AI Adversary Tactics
Real-world AI/ML attack techniques mapped from threat intelligence
ML Attack StagingAML.TA0001
  • ▸Acquire public ML artifacts
  • ▸Obtain capabilities via supply chain
  • ▸Stage malicious model weights
ML Model AccessAML.TA0002
  • ▸Inference API access
  • ▸Physical environment access
  • ▸ML model file access
ReconnaissanceAML.TA0000
  • ▸Search for victim AI assets
  • ▸Discover ML model outputs
  • ▸Identify data dependencies
Exfiltration via AIAML.TA0010
  • ▸Model inversion attack
  • ▸Membership inference
  • ▸Training data reconstruction
Prompt InjectionAML.T0051
  • ▸Direct prompt injection
  • ▸Indirect via retrieved data
  • ▸Jailbreak via role-play
Adversarial EvasionAML.T0015
  • ▸Craft adversarial examples
  • ▸Bypass content filters
  • ▸Perturbation attacks
Supply Chain CompromiseAML.T0010
  • ▸Poison training data
  • ▸Backdoor ML model
  • ▸Compromise model repository
Erode ML Model IntegrityAML.TA0005
  • ▸Inject backdoor
  • ▸Introduce bias
  • ▸Corrupt fine-tuning pipeline
Found a new AI vulnerability pattern?

Submit for review. Verified entries receive a VBS advisory ID and researcher credit.