Home · FAQ
FAQ
Questions buyers ask before they sign.
Straight answers on what to buy, how we test, where your data lives and who we are. Where something isn’t available yet, we say so.
Choosing and buying
Which offering should we start with?
Start from the moment you’re in. Shipping a new AI feature: an AI Launch Check maps what it can reach and what could go wrong before go-live. A deal, an audit or a big change to your AI coming up: an AI Red Team tests the live system and ends in signed evidence. Selling to enterprises every quarter: KorelexOne keeps your AI security and compliance posture current on every release. If you’re unsure, a 20-minute scoping call settles it.
How long does an engagement take?
An AI Red Team typically takes about 3 weeks from scoping to signed evidence. If a deal date is driving it, tell us when you scope.
How is pricing set?
Pricing shared on request. Every engagement is scoped to your system, so we quote after a 20-minute scoping call.
What do you need from us to start?
A 20-minute scoping call, a mutual NDA before scoping, and written authorization to test. We do not start testing without written authorization.
Will this help us pass a buyer’s security review?
That is what it is built for. Every finding is mapped to the OWASP LLM Top 10 and MITRE ATLAS, signed and dated, and comes with answers for the security questionnaire in front of you. We can also join the review as your independent AI security partner, speaking to our own findings.
How we test
Is it safe to test our production system?
Testing is read-only by default, with hard request caps, and PII is redacted at the chokepoint. We record what the system refuses through every legitimate channel before claiming a bypass.
How is this different from a pentest or a vulnerability scanner?
A pentest tests your application and infrastructure; a scanner matches known patterns. Neither treats the AI itself as the attack surface: what it reads, what it can reach and what it can be talked into. One platform in our case studies had been pentested and still had six criticals reachable from a free trial seat.
How do you avoid false or inflated findings?
Every finding is re-run verbatim before it is written down: same status, same leak, or it is not reported. Each carries exactly one label from our evidence standard, and the actual requests and transcripts ship with the report so you can verify any claim yourself.
Which frameworks do you map findings to?
The OWASP LLM Top 10, MITRE ATLAS, the EU AI Act, GDPR, the RBI advisory and India’s DPDPA. KorelexOne also maps your posture to SOC 2, ISO 27001 and ISO 42001.
Do you also sell the fix?
No. A tester who also sells the remedy has a reason to find something. We don’t build models and we don’t sell what we test: you get root-caused findings and fix guidance, then a retest. We verify controls; we never operate them.
Your data, hosting and AI
What deployment models does KorelexOne offer?
KorelexOne runs as SaaS today. Dedicated deployment options are on our roadmap; ask us about timing for your case.
Can KorelexOne run in our country or in our own cloud?
Not today. In-country deployment is planned; tell us your region and we’ll share where it stands.
Where does our data live today?
KorelexOne is hosted in Mumbai, India.
Does KorelexOne send our data to third-party AI models?
No. Customer information is redacted before any AI processing. No customer PII is kept for training; only anonymised, aggregate usage data is kept, to improve the product.
Who at Korelex can access our data?
Only authorised Korelex staff assigned to your account can access your data.
How do you handle the credentials we share for testing?
Credentials for an engagement are kept in a vault and removed when the engagement is complete.
Where are your testers based?
Our testers are based in India and test remotely.
The company
Who is behind Korelex?
Korelex is a product of Vaultmark Private Limited, based in Indiranagar, Bengaluru. It is founder-led: Manish Goyal architected Private AI Identity at VMware, built data governance for Microsoft OneDrive, and ran identity and data systems in production at Amazon, Microsoft and VMware. Meet the team.
Who are your clients?
AI-native companies selling into the enterprise, including Leap10x and SydanticIQ, who let us name them. Their AI platforms are trusted by some of India’s largest banks and enterprises. Read their stories in their own words.
Still deciding?
Ask us the question that isn’t here.
Twenty minutes is enough to scope it, and to answer what your security team will ask.