HomeNews

366K Scanner Requests Target Claude and Cursor AI Config

Devansh
Devansh
•
Nov 11, 2026 10:30 PM
•
0
 min read
Select Emergent as your Preferred news source
366K Scanner Requests Target Claude and Cursor AI Config

💡 TL;DR

  • Security researchers documented 366,845 automated scanner requests targeting claude.json and .cursor/mcp.json AI configuration files across monitored systems.
  • Mass scanning activity reveals attackers systematically probing for exposed AI assistant configuration files containing API keys and integration settings.
  • The analysis highlights emerging security risks as AI coding tools become widespread, with configuration files becoming high-value targets.

Security researchers at HoneyLabs have published an analysis of 366,845 automated scanner requests targeting AI assistant configuration files, specifically claude.json and .cursor/mcp.json endpoints. The research, officially released on November 11, 2026, reveals systematic probing activity as attackers hunt for exposed configuration files containing API keys, integration settings, and other sensitive data used by popular AI coding tools.

Mass Scanner Activity Documented

The analysis tracked scanner requests across multiple honeypot systems over several months, identifying distinct patterns in how attackers probe for AI configuration files. The claude.json file, used by Anthropic's Claude AI assistant, and .cursor/mcp.json, associated with Cursor's Model Context Protocol implementation, emerged as primary targets. Researchers observed automated bots systematically checking common file paths and directory structures where developers might store these configuration files.

Attack patterns showed scanners testing multiple URL variations, including root directories, hidden folders, and common configuration paths. The volume of requests indicates widespread automated scanning rather than targeted attacks, suggesting attackers are casting a wide net across internet-accessible systems.

Configuration File Risk Landscape

AI coding assistants increasingly rely on configuration files to manage API credentials, model preferences, and integration settings. The claude.json file typically contains Claude API keys, model selection parameters, and prompt templates. The MCP client configuration in .cursor/mcp.json can include server connections, authentication tokens, and tool permissions. When these files are accidentally exposed through misconfigured web servers or version control systems, they provide attackers direct access to AI services and development environments.

The research highlights how AI tool adoption has created new attack surfaces. Developers working with Claude, Cursor, and similar tools may inadvertently expose configuration files through improper .gitignore settings, misconfigured development servers, or cloud storage permissions. Each exposed file can grant unauthorized access to paid AI services, expose proprietary prompts, or reveal internal system architecture.

Scanner Methodology and Detection

HoneyLabs deployed specialized honeypots designed to log configuration file access attempts without exposing real credentials. The systems recorded request headers, IP addresses, and scanning patterns to build a comprehensive picture of attacker behavior. Analysis revealed scanners using common user agents, automated tooling signatures, and distributed IP addresses suggesting botnet activity.

Most scanning attempts followed predictable patterns: checking for files in web root directories, testing common subdirectories like /config or /.cursor, and probing for backup files with extensions like .bak or .old. Some sophisticated scanners checked multiple variations simultaneously, indicating automated tooling specifically designed for AI configuration file discovery.

Security Implications for AI Development

The scale of scanning activity underscores growing attacker interest in AI development infrastructure. As organizations adopt AI coding assistants, configuration management becomes a critical security consideration. Exposed API keys can lead to unauthorized service usage, resulting in unexpected costs and potential data exposure. Configuration files may also reveal internal tool usage patterns, integration architectures, and development workflows that attackers can exploit in targeted campaigns.

Security experts recommend treating AI configuration files with the same sensitivity as database credentials or cloud access keys. Best practices include storing configuration files outside web-accessible directories, using environment variables for sensitive values, and implementing robust access controls on development systems.

What This Means

The HoneyLabs analysis reveals a maturing threat landscape around AI development tools. With nearly 367,000 documented scanner requests targeting specific AI configuration files, the data shows attackers actively adapting to new technologies and hunting for emerging vulnerabilities. Organizations using Claude, Cursor, and similar tools must prioritize configuration security, implement proper file access controls, and monitor for unauthorized scanning activity. As AI coding assistants become standard development tools, configuration file security will remain a critical focus for both security teams and tool vendors.

About the writer

Devansh is a marketing enthusiast, keeps himself updated about the evolving startup ecosystem. He likes problem solving and does so by writing content; directly affecting reader's intent and decision making. Working at Emergent, he covers areas like AI workflows, use cases, automation and more. Interested in consumer psychology, he enjoys blending creativity with strategic thinking to provide impactful brand experiences.

Start Building
on Emergent today
Try Emergent