Guardrails
A browser-based safety tool that detects sensitive information before it reaches AI assistants, helping organisations protect personal data, strengthen responsible AI use, and improve prompts for more effective AI interactions.
The challenge
AI assistants have become part of everyday work across the social sector. Staff use them to draft emails, summarise reports, prepare funding applications, analyse documents, and support service delivery.
But these tools also introduce new risks.
Personal information, case notes, addresses, phone numbers, and other sensitive details can easily be copied into AI systems without users realising the implications. While policies and training are important, they often rely on people remembering guidance in the moment.
We wanted to build something that provides practical support exactly when it is needed.
Our approach
Guardrails works directly in the browser, helping people use AI more safely wherever they work online.
Instead of replacing existing AI tools, Guardrails works alongside them by reviewing prompts before they are submitted. It highlights potentially sensitive information, suggests safer alternatives, and helps users improve their prompts without interrupting their workflow.
The goal isn't to stop people using AI. It's to help them use it responsibly.
Key features
- Protect sensitive information
- Detect personal information, contact details, attachments, and other sensitive content before it is shared with AI tools.
- Improve prompts
- Help staff write clearer, more structured prompts that produce better AI responses while reducing unnecessary risk.
- Support organisational guidance
- Bring organisational AI policies into everyday work by providing guidance and resources exactly when people need them.
- Work with existing AI tools
- Guardrails integrates into the tools teams already use, supporting platforms such as ChatGPT, Claude, and Gemini without requiring organisations to change their workflow.