Who This Topic Is For
Agencies and owners who can already open a client bot or a public model chat and are allowed to test that address.
Documentation
AI Security
Point a credit-based check at a live chatbot you can already open. Read the reply, the result, and the fix.
Prerequisites
Before You Start
Use this checklist to make sure the workflow guidance applies cleanly to your current task.
- You can sign in and open AI Security.
- You can already open the chatbot, either at a direct address or in a page chat box.
- You are allowed to test that exact address. An API key is not enough for a private client bot.
Expectations
What To Expect
Use this section to set the right outcome before you start the workflow.
A flat credit charge
Baseline is 12 credits. Standard is 24 credits. A practice fixture on your own machine is not charged. A longer profile stays hidden until it is turned on.
No real side effect
Email, payment, and data-change checks use a Vulnify sink or are blocked before they run. A real charge is never placed. The report still says whether the bot tried, listed, or only claimed that ability.
Playbook
Step-By-Step Guidance
Follow these steps in order for a reliable and repeatable outcome.
Open AI Security.
Sign in and go to /ai-security. The public explanation stays at /ai-security-check.
Add the address and confirm the test.
Paste the direct chat address or the page that holds the chat box. Confirm you are allowed to test it.
A planted secret is optional.
Start Baseline or Standard.
Baseline is the short run.
Standard covers more cases. Read the reply, the result, and the fix.
Run the same check again after a change.
Examples
Worked Examples
These scenarios show how the workflow looks in practice, including the result you should see.
A client support chat
The agency can already open the client bot. They add that page, confirm permission, and run Standard.
A planted secret returns
The team adds one test secret the bot should never repeat. The reply includes it.
A payment sentence with no call
The bot says it can charge a card and does not call a payment tool.
Validation
Validation Checklist
Use this checklist to confirm the workflow was completed correctly.
- The address is one the tester can already open.
- Authorization was confirmed for that address.
- Baseline was charged 12 credits, or Standard was charged 24, unless the target was a local fixture.
- A quiet result says no violation was observed in these tests.
- No sentence says the bot is safe or certified.
- Email, payment, and data checks did not reach a real person, card, or record.
Troubleshooting
Common Problems And Fixes
If something does not match expectation, check these common failure modes first.
The run could not speak to the bot
The address did not return JSON or streamed text, or the page chat box could not be used. An unsupported run refunds the credits. A finished run that observed no violation does not.
I expected a safe badge
A quiet result means these tests did not show a problem. It is not a certification and it does not cover every trick.
The bot claimed it could email someone
A claim is recorded as likely. Confirmed requires a caught or blocked attempt, or a planted secret or marker in the reply.
Navigate
Related Pages
Use these links to continue your workflow without losing context.
FAQ
AI Security FAQs
Common questions for this topic.
No. Use a chatbot you can already open. Vulnify does not bypass a login.
Next Recommended Action
Continue to the best next page based on where you are in your workflow.