Everything Emil catches before it leaks

Emil sits between you and the AI tools you use — screening every prompt for PII, secrets and client data, and every output for content that shouldn’t ship. One free extension for your browser, one line of config for your code.

What it detects

🪪

PII

SSNs, names, dates of birth, addresses, phone numbers, email addresses — caught and redacted before they reach the model.

🔑

Secrets & credentials

API keys, passwords, tokens, connection strings. The stuff that ends up in a breach report when it leaks.

📁

Client & business data

Client records, contracts, financials — the confidential content your engagement letters promise to protect.

💉

Prompt injection

Hidden instructions in pasted text that try to hijack the model. Detected and blocked before they run.

🚫

Unsafe output

For AI builders: 14-category safety screening on what your model generates, with an explicit-content backstop.

📊

Risk scoring

Every prompt scored across five axes — safety, privacy, security and more — so policies can warn, redact, or block by severity.

For everyone

The browser extension

Install once, protected on every AI chat. Free.

  • Screens ChatGPT, Claude, Gemini, Copilot, Perplexity and DeepSeek automatically
  • Runs locally — sensitive text never leaves your browser
  • Pauses the send and offers one-click redaction
  • No account, no API key, no configuration
Learn more about the extension →
For developers

The proxy API

Guardrails for the AI products you’re building.

  • OpenAI-compatible: change one line (the base URL) and you're covered
  • Screens input and output — prompts in, generations out
  • Fail-closed with a full audit trail
  • Works with Cursor, Claude Code, Copilot, Lovable, v0, Ollama and self-hosted models
See the install guide →

Built for regulated work

Detection categories are mapped to the frameworks your auditors ask about — HIPAA, the EU AI Act, FERPA, Circular 230. Teams can force-install the extension via the Google Admin Console or MDM, push central policies, and see which AI tools their org actually uses.

The next paste is the one that leaks.

Thirty seconds to install. Free.