SafePaste — AI Paste Shield

Warns you when pasted text may manipulate AI behavior (prompt injection). Runs locally with no data collection.

SafePaste — AI Paste Shield SafePaste warns you when pasted text may manipulate AI behavior (prompt injection). This helps reduce risk when copying text from websites, documents, PDFs, or text copied from OCR’d screenshots into AI chat tools. Features - Detects paste events in supported AI chat inputs - Flags common prompt-injection patterns (e.g., “ignore previous instructions”) - Simple warning with “Paste anyway” or “Cancel” - Optional strict mode and warning threshold controls - Local-only: no tracking, no accounts, no data collection Supported sites (v1.0) - ChatGPT (chatgpt.com + chat.openai.com) - Claude (claude.ai) - Gemini (gemini.google.com) - Copilot Chat (copilot.microsoft.com) - Groq Chat (chat.groq.com) - GroqCloud Console/Playground (console.groq.com) Privacy SafePaste runs entirely on your device. It does not send pasted text anywhere. SafePaste analyzes pasted text only. It does not analyze uploaded images. Notes SafePaste is a safety aid, not a guarantee. It may miss attacks or flag harmless text.