What is an AI firewall?
An AI firewall is a control point between people or applications and AI models. It checks prompts on the way in and answers on the way out, masking sensitive data, stopping prompt injection and flagging unsafe content, so that one policy covers every AI tool without blocking its use.
How does an AI firewall work?
-
Sit in the path
A browser extension, the endpoint agent or an AI gateway sees the traffic.
-
Check the prompt
Personal data is masked; secrets and source code are caught.
-
Check the answer
Masked values are restored; sensitive data and unsafe links are flagged.
-
Record
Each decision is kept as evidence.
What does an AI firewall decision look like?
- Data
- Support ticket holding an IBAN
- User
- Support agent in Zendesk
- Destination
- LLM behind the AI gateway
- Policy
- Bank details: mask before sending
- Decision
- Record IBAN masked, restored in the answer
Where is the AI firewall in GuardStance?
AI Firewall works in the browser, on the desktop and between your apps and LLMs.