Unbound
The Claude Code Extension Backdoor in Cursor's Marketplace: What Happened and What to Do

Supply Chain

Cursor

Claude Code

IDE Extensions

Shadow AI

Incident Analysis

The Claude Code Extension Backdoor in Cursor's Marketplace: What Happened and What to Do

Cursor's marketplace served malicious Claude Code extension builds 2.1.312 and 2.1.313. Check for svc-telemetry in authorized_keys; uninstalling leaves the backdoor in place.

Vignesh SubbiahVignesh Subbiah13 Min Read

Latest Blogs

Break Out to Finish the Task
AI Security30 Jul, 2026•9 Min Read

Break Out to Finish the Task

Agentic systems are increasingly optimized to get the job done even when that means pushing past the guardrails designed to contain them.

Nanda Pranesh SNanda Pranesh S
The Claude Desktop App Governance Playbook
Claude21 May, 2026•13 Min Read

The Claude Desktop App Governance Playbook

Claude Desktop is no longer a chat app. It hosts chat, Cowork, and a launch surface for Claude Code, with MCP support, file access, and agentic execution on every knowledge worker's machine. Here is the governance playbook for security teams running it at scale.

Rajaram SrinivasanRajaram Srinivasan
Unbound Brings Full Enforcement to GitHub Copilot
AASB19 May, 2026•8 Min Read

Unbound Brings Full Enforcement to GitHub Copilot

GitHub shipped postToolUse hooks in the Copilot CLI on May 18, 2026. Unbound now governs Copilot at parity with Claude Code, Codex, and Cursor: audit, warn, approve, block on the most widely deployed AI coding agent in the enterprise.

Rajaram SrinivasanRajaram Srinivasan
Prompt Injection in Coding Agents: Every Attack, Every Defense
AI Security29 Apr, 2026•9 Min Read

Prompt Injection in Coding Agents: Every Attack, Every Defense

Every prompt injection vector specific to AI coding agents, with proof-of-concept patterns, detection signals, and defensive controls. Why action policy beats detection.

Rajaram SrinivasanRajaram Srinivasan
Top MCP Server Risks in Production: A Red Team Walkthrough
MCP27 Apr, 2026•11 Min Read

Top MCP Server Risks in Production: A Red Team Walkthrough

Three MCP exploit chains end to end: tool poisoning, exfil chain, and confused deputy. Detection telemetry, controls, and AASB mapping for each.

Rajaram SrinivasanRajaram Srinivasan
Every Known MCP Attack Pattern, Mapped: A Defender's Taxonomy
MCP20 Apr, 2026•10 Min Read

Every Known MCP Attack Pattern, Mapped: A Defender's Taxonomy

The five attack classes targeting AI coding agents through Model Context Protocol servers. Preconditions, mechanism, detection signal, and the controls that stop each one.

Rajaram SrinivasanRajaram Srinivasan
Governing Claude Across Web, Desktop, and Code: What Security Teams Need to Know
Claude14 Apr, 2026•7 Min Read

Governing Claude Across Web, Desktop, and Code: What Security Teams Need to Know

Claude operates across web, desktop, CLI, and Cowork surfaces. Each has a different risk profile. Learn how to govern Claude usage consistently across your organization.

Rajaram SrinivasanRajaram Srinivasan
Shadow AI Coding Agents: The Security Risk Your Engineering Team Isn't Talking About
Shadow AI5 Apr, 2026•5 Min Read

Shadow AI Coding Agents: The Security Risk Your Engineering Team Isn't Talking About

Shadow AI coding agents are running in your development environment without security approval. Learn how to detect them, assess the risk, and build governance that scales.

Rajaram SrinivasanRajaram Srinivasan
Securing Cursor, Codex, and Claude Code: A Comparison of AI Coding Agent Risks
Cursor31 Mar, 2026•4 Min Read

Securing Cursor, Codex, and Claude Code: A Comparison of AI Coding Agent Risks

Compare the security architectures and risk profiles of Cursor, Codex, and Claude Code. Understand what each tool accesses and how to govern them with one policy engine.

Rajaram SrinivasanRajaram Srinivasan
What Happens When an AI Coding Agent Goes Rogue: Real-World Incident Scenarios
AI Security25 Mar, 2026•5 Min Read

What Happens When an AI Coding Agent Goes Rogue: Real-World Incident Scenarios

Four realistic scenarios showing how AI coding agents can leak credentials, exfiltrate data, and compromise infrastructure. Each maps to a documented 2025 incident and a prevention path.

Rajaram SrinivasanRajaram Srinivasan
AI Coding Agent Security for Financial Services: Compliance Requirements and Controls
Financial Services23 Mar, 2026•4 Min Read

AI Coding Agent Security for Financial Services: Compliance Requirements and Controls

AI coding agents in financial services create regulatory exposure under OCC, FFIEC, SEC, and NY DFS guidance. Here are the specific risks and required controls.

Rajaram SrinivasanRajaram Srinivasan
The CISO's Guide to AI Coding Agent Risk
CISO20 Mar, 2026•6 Min Read

The CISO's Guide to AI Coding Agent Risk

A practical framework for CISOs managing AI coding agent risk. Covers the five risk categories, where existing tools fall short, and how to build a governance program.

Rajaram SrinivasanRajaram Srinivasan
The State of AI Coding Agent Risk
AASB20 Mar, 2026•14 Min Read

The State of AI Coding Agent Risk

A comprehensive analysis of the security, compliance, and operational risks enterprises face as AI coding agents become the default developer interface — mapped to the OWASP Top 10 for Agentic Applications.

Rajaram SrinivasanRajaram Srinivasan
AWS Kiro Didn't Just Delete an Environment. It Exposed the Missing Control Plane for AI Coding Agents.
AASB18 Mar, 2026•8 Min Read

AWS Kiro Didn't Just Delete an Environment. It Exposed the Missing Control Plane for AI Coding Agents.

A reported December 2025 AWS incident involving the Kiro AI coding agent turned a small software fix into a 13-hour outage. The strategic lesson: enterprises need a dedicated governance layer between AI coding agents and the systems they can change.

Rajaram SrinivasanRajaram Srinivasan
How to Govern AI Coding Agents Without Killing Productivity
AASB16 Mar, 2026•7 Min Read

How to Govern AI Coding Agents Without Killing Productivity

You do not need to choose between speed and control. The right governance model lets developers keep the AI coding tools that make them faster while applying discovery, posture checks, runtime controls, and approvals to the actions that create enterprise risk.

Rajaram SrinivasanRajaram Srinivasan
Your Team Adopted Cursor. Do You Know Which MCP Servers It Can Reach?
MCP8 Mar, 2026•4 Min Read

Your Team Adopted Cursor. Do You Know Which MCP Servers It Can Reach?

MCP servers give AI coding agents access to databases, APIs, and infrastructure. Most security teams have no idea they exist.

Rajaram SrinivasanRajaram Srinivasan

Start with a Free Two-Week Assessment

See every agent, MCP server and skill on your fleet, where the spend goes, and which actions to stop. Most teams find something unexpected in the first week.

Free for teams with 50 or more developers using coding agents.