Shadow Research
AI privacy research
What AI coding assistants and chat tools can expose, measured and documented. Analysis and practical guidance from the Shadow team, with sources and caveats.
Latest research
What AI coding tools send home in telemetry, and how to turn it off
What telemetry Claude Code, Codex, Gemini CLI, Copilot, Cursor, Cline, Continue, Aider and Zed collect, and the exact opt-out settings.
Where AI tools store MCP server credentials: a plaintext secrets audit
Where 9 MCP clients keep server API keys and OAuth tokens: config paths, project files, env var syntax, and which use the OS keychain.
Where do AI coding agents store your conversations on disk, and do pasted secrets persist in plaintext?
Audit of AI coding agent local storage: where sessions are saved, file format, retention defaults, and whether pasted secrets persist.
Do AI coding agents keep your .env files out of the prompt? A source-code and documentation audit
Source-code audit of AI coding agents: what controls stop .env files and secrets from reaching models, and which tools protect you by default.
AI Leak Report #1: 122,255 sensitive values caught on one developer laptop
In 30 days Shadow masked 122,255 sensitive values in AI traffic from one developer laptop, including 287 secrets and 1,405 emails. What it found and why.
Does Cursor train on my code? Privacy Mode, retention and what still leaves
Only if Privacy Mode is off. With it on, Cursor and its model providers do not train on your code, but every prompt still goes through Cursor's servers.
Private ChatGPT: how to use AI without sharing your data
A private ChatGPT keeps prompts out of training data and away from the provider. Compare settings, enterprise plans, self-hosted models and privacy proxies.
Is ChatGPT safe for company data? What happens to your prompts
Whether ChatGPT is safe for work data depends on the plan: Free, Plus and Pro can train on chats by default; Business, Enterprise and the API do not.
What is an LLM proxy? How it works and when you need one
An LLM proxy sits between your apps and AI providers to route, log, limit or protect requests. The types, how they differ, and how to pick one.
What is shadow AI? Risks, examples and how to detect it
Shadow AI is AI use your IT team doesn't know about. What it is, the risks, real examples and how to detect it without blocking work.
Free AI acceptable use policy template, with enforcement
A free AI acceptable use policy template for companies, with the controls that make each rule enforceable rather than a document nobody reads.
AI Privacy
Research on data privacy risks when using AI tools, what happens to your prompts, and how to protect sensitive information.
- Private ChatGPT: how to use AI without sharing your data
- Is ChatGPT safe for company data? What happens to your prompts
AI Coding Tools
Privacy and security analysis of AI coding assistants like Cursor, GitHub Copilot, Claude Code, and other developer AI tools.
- What AI coding tools send home in telemetry, and how to turn it off
- Where AI tools store MCP server credentials: a plaintext secrets audit
- Where do AI coding agents store your conversations on disk, and do pasted secrets persist in plaintext?
- Do AI coding agents keep your .env files out of the prompt? A source-code and documentation audit
- Does Cursor train on my code? Privacy Mode, retention and what still leaves
Enterprise AI
How businesses can use AI safely: enterprise plans, data agreements, compliance considerations, and governance frameworks.
- Private ChatGPT: how to use AI without sharing your data
- Is ChatGPT safe for company data? What happens to your prompts
- Free AI acceptable use policy template, with enforcement
Shadow AI
Understanding and detecting unauthorized AI use in organizations, and how to govern AI without blocking productivity.
- What is shadow AI? Risks, examples and how to detect it
- Free AI acceptable use policy template, with enforcement
AI Security
Security risks of AI tools including data leaks, credential exposure, and how to protect against them.
- What AI coding tools send home in telemetry, and how to turn it off
- Where AI tools store MCP server credentials: a plaintext secrets audit
- Where do AI coding agents store your conversations on disk, and do pasted secrets persist in plaintext?
- Do AI coding agents keep your .env files out of the prompt? A source-code and documentation audit
- AI Leak Report #1: 122,255 sensitive values caught on one developer laptop
- What is an LLM proxy? How it works and when you need one
- What is shadow AI? Risks, examples and how to detect it
Data Privacy
How personal data, customer information, and sensitive business data can be exposed through AI tools.
- AI Leak Report #1: 122,255 sensitive values caught on one developer laptop
- Is ChatGPT safe for company data? What happens to your prompts
AI Governance
Policies, frameworks, and technical controls for governing AI use in organizations.
- What is shadow AI? Risks, examples and how to detect it
- Free AI acceptable use policy template, with enforcement
Developer Tools
AI tools for software development and how to use them without exposing code, secrets, or customer data.
- What AI coding tools send home in telemetry, and how to turn it off
- Where AI tools store MCP server credentials: a plaintext secrets audit
- Where do AI coding agents store your conversations on disk, and do pasted secrets persist in plaintext?
- Do AI coding agents keep your .env files out of the prompt? A source-code and documentation audit
- AI Leak Report #1: 122,255 sensitive values caught on one developer laptop
- Does Cursor train on my code? Privacy Mode, retention and what still leaves
- What is an LLM proxy? How it works and when you need one
Privacy Controls
Technical and administrative controls for protecting data when using AI: proxies, masking, enterprise settings.
- Does Cursor train on my code? Privacy Mode, retention and what still leaves
- Private ChatGPT: how to use AI without sharing your data
- What is an LLM proxy? How it works and when you need one
Guides and use cases
Step-by-step setups for the risks the research describes.
- Protect secrets from AI coding assistants
- Stop PII leaks to ChatGPT
- Anonymize prompts for Claude
- How Shadow masks and restores data
- Compare AI data-leak tools
Frequently asked questions
What does Shadow Research cover?
Evidence-based analysis of AI data privacy: what tools such as ChatGPT, Cursor and Claude Code send to model providers, what their privacy settings do and do not control, and practical ways to keep sensitive data out of prompts.
How can I protect my data when using AI coding assistants?
Keep secrets out of the workspace, turn on privacy modes where they exist, and route traffic through a masking proxy such as Shadow, which replaces sensitive values before a prompt leaves your machine. See Does Cursor train on my code?.
Does ChatGPT train on my company data?
Free, Plus and Pro chats can be used for training unless you opt out; Business, Enterprise and the API are not used for training by default. See Is ChatGPT safe for company data?.
What is shadow AI and why does it matter?
Shadow AI is AI use at work that IT does not know about: personal ChatGPT accounts, browser extensions, coding agents with personal keys. Company and customer data can reach providers with no contract or visibility. See What is shadow AI?.