diff --git a/docs/images/debug-mode.mp4 b/docs/images/debug-mode.mp4
new file mode 100644
index 000000000..165ffbb4d
Binary files /dev/null and b/docs/images/debug-mode.mp4 differ
diff --git a/docs/images/filter.mp4 b/docs/images/filter.mp4
new file mode 100644
index 000000000..913a63d7f
Binary files /dev/null and b/docs/images/filter.mp4 differ
diff --git a/docs/images/inspect-mode-message.png b/docs/images/inspect-mode-message.png
new file mode 100644
index 000000000..af11f2624
Binary files /dev/null and b/docs/images/inspect-mode-message.png differ
diff --git a/docs/mint.json b/docs/mint.json
index 7578adab4..7b1326dd0 100644
--- a/docs/mint.json
+++ b/docs/mint.json
@@ -131,6 +131,11 @@
"icon": "envelope",
"pages": ["api-reference/email/email_api"]
},
+ {
+ "group": "AI Safety",
+ "icon": "shield-check",
+ "pages": ["safety/introduction", "safety/inspect-mode", "safety/debug-mode"]
+ },
{
"group": "Resources",
"pages": [
diff --git a/docs/safety/debug-mode.mdx b/docs/safety/debug-mode.mdx
new file mode 100644
index 000000000..1fdd96fa8
--- /dev/null
+++ b/docs/safety/debug-mode.mdx
@@ -0,0 +1,45 @@
+---
+title: "Debug Mode"
+icon: "circle-info"
+description: "Understand why the AI said what it said."
+---
+
+When the AI provides an answer, there’s usually a reason behind it. This feature allows you to see the underlying data the AI used to form its response.
+
+## How to use this feature?
+
+
+
+ Filter sessions by the AI agent. This will display a list of all ongoing conversations where the AI is active.
+
+
+
+ Clicking this will open a popup showing the data the AI used to answer the question.
+
+
+
+ - **Actions**: These represent the steps the AI took to formulate its response.
+ - **Knowledge**: This shows the data or knowledge the AI relied on to answer the question.
+ - **Keywords**: These are the search terms or keywords the AI used in its hybrid search process.
+
+
+
+## How to enable debug mode on the widget?
+For development purposes, you can enable debug mode on the widget by passing the `debug` prop to the `initOpenScript` function. This way, the widget will log all the messages and events to the console and to the chat ui (default: false)
+
+
+Read more here
diff --git a/docs/safety/inspect-mode.mdx b/docs/safety/inspect-mode.mdx
new file mode 100644
index 000000000..c27134b2c
--- /dev/null
+++ b/docs/safety/inspect-mode.mdx
@@ -0,0 +1,30 @@
+---
+title: "Inspect Mode"
+icon: "eye"
+description: "See what the AI is doing in real time and intervene if needed."
+---
+
+This feature allows human agents to monitor the AI in real time. You can observe the AI’s actions, see what it's doing, and understand its reasoning.
+
+## How to use this feature?
+
+
+
+ Filter sessions by the AI agent. This will provide a list of all ongoing conversations where the AI is active.
+
+
+
+ Click on the session you want to inspect. This will open Inspect Mode.
+ 
+
+
+ If necessary, you can take control of the conversation by typing a message in the chat window. This will pause the AI and let you take over.
+
+
\ No newline at end of file
diff --git a/docs/safety/introduction.mdx b/docs/safety/introduction.mdx
new file mode 100644
index 000000000..5386a1ff6
--- /dev/null
+++ b/docs/safety/introduction.mdx
@@ -0,0 +1,68 @@
+---
+title: "Introduction"
+icon: "shield-check"
+description: "We have gone further than anyone in on the market to make sure that our AI is safe and reliable."
+---
+We know that AI, in its current form, is still new to the market. We’re constantly working to make sure it’s safe and reliable. That’s why we’ve implemented several measures to ensure the trustworthiness of our AI, giving you confidence when using it.
+
+## Observability & Interpretability
+
+These features are designed to give human agents the ability to see and understand the AI while it’s at work.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+## Assist Mode (formerly Draft Mode)
+
+For heavily regulated industries, we’ve introduced Assist Mode. This feature allows human agents to review the AI’s plan before sending it to the customer. Think of it as a human supervision layer for AI.
+
+
+
+
+
+
+## Compliance
+
+We are in the process of achieving SOC-2 certification, which is expected to be completed by Q4 2024.
\ No newline at end of file