BREAKING
macOS.Gaslight Targets AI Analysis
0
fake errors
0
KB
injected block
How the Prompt Injection Works
1
Embed fake errors
↓
2
LLM ingests text
↓
3
Agent aborts scan
Old Tactics vs Gaslight
AMOS
●
Sandbox evasion
●
Fake user dialogs
Gaslight
NEW
●
Targets AI tooling
●
38 chained messages
Rust Backdoor via Telegram C2
AI Triage Faces New Risk
AI NEWS BLITZ
SentinelLABS has uncovered a new macOS malware built to deceive AI-powered analysis tools.