I will do llm penetration testing and ai chatbot security testing


About this gig
Shipping an AI chatbot or agent? I do LLM penetration testing for prompt injection, jailbreaks and data leaks, before attackers or users find them.
WHAT I TEST
- Direct and indirect prompt injection
- Jailbreaks and system prompt extraction
- Sensitive data leaks from RAG and connected data
- Tool and function calling abuse (excessive agency)
- Unsafe output handling that leads to XSS or code execution
- OWASP Top 10 for LLM Applications (2025)
MY PROCESS
- You confirm scope, model provider and test access
- I map prompts, tools, data sources and roles
- Manual adversarial testing plus a structured attack library
- Report, then a free retest after you fix
WHAT YOU GET
- PDF report with severity, exact prompts, model responses and fix steps
- Guardrail and system prompt hardening advice
- Findings mapped to the OWASP LLM Top 10
- Retest report
WHY ME
I test AI products and the web apps and APIs around them, so I catch the attack paths that cross both.
I only test AI systems you own or are authorized to test. Tell me what your AI can access for a fixed quote.
Get to know Faizan U.
Web App and API Penetration Tester for SaaS and AI Products
- FromPakistan
- Member sinceNov 2026
- Avg. response time4 hours
Languages
English, Arabic
FAQ
Is this a manual pentest or just an automated scan?
Manual. I run a scan only to map the attack surface, then test every role and flow by hand in Burp Suite. Scanners miss access control and business logic flaws, the findings that matter most for SaaS apps.
Will testing break my app or affect real users?
I avoid destructive tests by default and never run denial-of-service attacks. Staging is preferred. If you only have production, we agree a testing window and I throttle traffic.
Will this report satisfy my SOC 2 or ISO 27001 auditor?
Most auditors accept an independent pentest report with scope, method, findings and retest results, which is what you get. I am not an audit firm and do not issue certificates, so confirm scope with your auditor first.
What do you need from me to start?
Written confirmation that you own or are authorized to test the target, the URL or API docs, one test account per role, and anything out of scope. The requirements form asks for all of it.
Do you need my source code?
No. This is a grey-box test using test accounts. If you also want a code-level security review, message me for a custom offer.
What if you find nothing serious?
You still get a full report: what was tested, how, and evidence that each control held. That documentation is what customers and auditors ask for.
Is the retest really free?
Yes, within 14 days on Basic and 30 days on Standard and Premium. I verify each fix and update the report to show what is closed.
Will you keep my data confidential?
Yes. Test data stays on my encrypted machine and is deleted 30 days after delivery. Findings are never shared, and I will sign your NDA on request.
Which AI models and platforms do you test?
Any model your app uses: OpenAI, Anthropic, Gemini, open-source models or custom agents. I test your application layer, not the provider's model itself.
Will you attack the AI provider's systems?
No. I test only your app, prompts, tools and data connections, within your provider's usage policies.

