LLM applications
Chat, search, summarisation and copilots built on top of a language model.
- Prompt injection, direct and indirect
- Jailbreaks and guardrail bypass
- Sensitive data and system prompt leakage
- Unsafe handling of model output
I am Santhosh Tuppad, an independent ethical hacker. I test LLM applications, AI agents and chatbots for the weaknesses that turn into data leaks, abused tools and lost trust.
Models do not fail like normal software. They fail through language, context and the tools you hand them. I test the whole system around the model, not just the model.
Chat, search, summarisation and copilots built on top of a language model.
Systems that call tools, browse, run code and act on behalf of users.
Customer facing assistants wired to your documents, data and APIs.
The AI is only one part. The app, API and mobile layers carry real risk too.
I started at 16, driven by plain curiosity, and grew into a white hat ethical hacker. Over the years I co-founded two testing companies built on one belief: testing is a deeply intellectual, exploratory craft, not a checklist.
Today I work as an independent consultant. I pair my own experience with AI agents, guiding them to surface bugs and to help produce fixes too. It makes me faster and sharper without losing the human judgement that finds what tools miss.
At heart I am a privacy fighter. Setting security standards that protect ordinary users matters more to me than any client logo.
Self taught, curious, and hooked on how systems break.
A testing company built around exploratory, thinking testers.
Training and community work that spread the craft of testing.
Security and quality work under my own banner.
Security testing, usability review and AI evaluation for product teams.
Securing LLMs, AI agents and LLM based applications, with OWASP Top 10 for LLMs as the compass.
Methods I work with include PTES, NIST and OWASP guidance, plus exploratory testing and the SELEUM usability heuristic.
Speaking at taqelah! Testing Conference 2027, Matrix @ Biopolis, Singapore.
I do not only break software. I build small, privacy-first tools and share the code, so you can read how I think before we ever speak.
Hides your real password length from anyone looking over your shoulder. The dots on screen do not match what you type, and the password itself is left untouched. No account, no data collected.
Chrome Web Store (opens in a new tab)Teaches you one Chrome or operating system keyboard shortcut at a time, in a small floating card. It collects no browsing data and no personal information.
Chrome Web Store (opens in a new tab)An open source testing browser for exploratory testing by humans and AI, created by Rikard Edgren. I contribute bug reports and fixes.
View on GitHub (opens in a new tab)The source for my own tools is on GitHub (opens in a new tab).
Within 2 weeks, we had a powerful report with 20 plus issues including cart manipulation, back end security gaps and more. Santhosh also gave clues to developers on root cause of issue and best practices to implement.
Santhosh is the person who has become the most preferred External VAPT vendor in a short time for us. … I recommend his VAPT services to any organisation that wants to secure its applications.
If India had an Olympic software testing team, Santhosh would be its captain. He’s a remarkably energetic, inventive thinker, who pursues the truth with single-minded integrity.
Since then he’s gone from strength to strength, in particular in the field of security and ethical hacking. … When I have a question on software security, Santhosh is the first person I contact.
Santhosh is one of the best testers I’ve ever worked with. … He’s a true software testing pro, a master at security and exploratory testing.
Excerpts from public recommendations on LinkedIn (opens in a new tab).
Before any paid work, I offer a free trial so you can experience my style of working: how I think, how I communicate and how I fit with your team. How long it runs depends on your product, so we agree on that together after our first conversation. It is not a bug hunt with a scorecard. Vulnerabilities may turn up along the way, but they are not the point.
A conversation so I understand your application and its context.
I suggest a trial length that suits your product. We sign an NDA before I touch anything.
How I explore, take notes, explain risk in plain language and talk with your developers.
If the style fits, we continue, short term or ongoing. If not, there is no obligation.
Remote by default. Tell me what you are building and I will reply personally.
Book an appointment to schedule a call. Pick a time that suits you for a short intro call, and we can plan your free trial from there.
Schedule a callPrefer to write directly?
ondsectest@gmail.comOr connect on LinkedIn (opens in a new tab) to see what I am building and sharing.
Every engagement starts with a signed NDA.