Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Humans have always sought to streamline productivity through the most convenient solutions available, prioritizing speed to stay ahead and gain an edge over the competition. From the assembly line to the cloud, the goal remains the same: do more with less friction. Today, that convenience is synonymous with AI. While these tools have revolutionized how we work, the reality remains that rapid innovation always comes with a hidden cost.

Model Context Protocol (MCP) servers facilitate the integration of third-party services with AI applications, but these benefits come with significant risks. If a trusted MCP server is hijacked or spoofed by an attacker, it becomes a dangerous vector for prompt injection and other malicious activities. One way attackers infiltrate software supply chains is through brand impersonation, also known as typosquatting—creating malicious resources that closely resemble trusted ones.

What permissions are developers granting to Claude Code, and could those permissions pose a risk if the coding agent were exposed to malicious inputs? To answer this question, we turned to GitHub, the website where developers go to share their private configuration files. From Github we collected a dataset of 18,470.claude/settings.local.json files, each containing the permissions that a user granted to Claude Code for a software project.

Organizations face a complex cybersecurity conundrum. Attack surfaces are expanding faster than SOC teams can scan. All of which is leading to a never-ending cycle of swivel-chair security, context-free lists, increased alert fatigue, and slow remediation. The strategic pivot needed to combat this is Continuous Threat Exposure Management (CTEM). A structured and essential alternative that moves teams away from reactive scanning to proactive, ongoing validation and prioritization.