<defaltam/>

How it works

Website chatbot security basics

Prompt injection, data leakage, abuse and rate limits — the security questions to ask before putting an AI chatbot on a public website.

In short

The main risks of a public chatbot are prompt injection, leaking information you did not mean to publish, and abuse of the message allowance.

Keep the knowledge base to public content, apply rate limits, and never give the bot access to systems it does not need.

4.7·130 reviews

Key facts

Key facts

Do not index what you would not publish

The simplest security rule for a website chatbot is that its knowledge base and your public site should contain the same information. Internal pricing sheets, staff notes and draft pages do not belong in it.

Limit what it can do

A chatbot that only reads content and captures enquiries has a very small blast radius. Every integration you add — order lookups, account changes — expands it, so add them deliberately and with verification.

Expect people to test it

  • · Someone will ask it to ignore its instructions.
  • · Someone will paste a long block of text pretending to be a system message.
  • · Someone will spam it to burn your allowance.
  • · Grounding, refusal instructions and rate limits handle all three.

The product behind this guide

A chatbot trained on your own website, live today

Defaltam builds AI chatbots that read your public pages and answer with your real prices, services and policies. Support is €79/month, Sales is €149/month, both start with a 5-day free trial, and installation is one script tag.

FAQ

Questions people ask about this

Is it safe to put an AI chatbot on a public website?

Yes, provided its knowledge base contains only content you would publish anyway, it has rate limits, and it has no access to systems beyond answering and capturing enquiries.

What is prompt injection?

A visitor writing text designed to override the bot's instructions. A well-configured bot treats visitor messages as questions, not as instructions about how to behave.

Can someone extract my private data?

Only what you indexed. Keep internal documents out of the knowledge base and there is nothing private to extract.

How is abuse handled?

Message and usage limits per session, which also protect you from runaway costs.

Related

Keep reading.