AI Threat Analysis Reports (AI TARs)

Weekly intelligence on AI-native software supply chain threats, vulnerabilities, and emerging attack patterns.

How One Agent's Redefined Success Became Hugging Face's Breach

Reasoning and Goal Integrity

GhostCommit: Hidden in Plain Sight, Read by Your Agent

Input and Instruction Manipulation

Friendly Fire: When Defensive AI Agents Become the Execution Path

Supply Chain Threats

MCP Tool Poisoning: The Tool Didn't Change. Its Instructions Did.

Supply Chain Threats

macOS.Gaslight Turns the Analyst’s AI Against the Analysis

Input and Instruction Manipulation

How Shai-Hulud Poisoned the AI Development Pipeline

Supply Chain Threats

Your AI Is Exposed and Talking to Strangers

Authorization and Identity

Helpdesk Hijack: Persuasion as Permission

Authorization and Identity

When the Comment Became the Control

Reasoning and Goal Integrity

Agentic AI Skills Marketplace Risk

Authorization and Identity

Vibe Coding at Your Own Risk

Authorization and Identity

The Obedient Accomplice

Instruction Manipulation

When the Tool Pulls Its Own Trigger: Cursor AI's Autonomous Production Wipe

Reasoning and Goal Integrity

NomShub Attack: AI Agent Prompt Injection Leading to Remote Code Execution

Instruction Manipulation

The Single-Click Microsoft Copilot Attack

Input and Instruction Manipulation

Reasoning Hijack Leading to Authorization Drift

Reasoning and Goal Integrity

Subscribe to Lineaje AI Threat Lab Intelligence