Test
Red-team LLMs, RAG applications and AI agents; document reproducible findings; verify that corrective actions address the underlying issue.
As enterprise AI moves into customer decisions, business processes and autonomous workflows, security checks and audit evidence have to keep pace. Cygeniq brings testing, governance and evidence preparation together in one continuous service, aligned to the CERT-In AI Security Auditing Guidelines v1.0.
For risk and security leaders, the issue is not simply whether AI passed a test before release. It is whether the organisation can explain what is running, where the risks sit and what changed since the last review.
Why now: CERT-In published its AI Security Auditing Guidelines v1.0 on 4 November 2026. The guidelines define risk tiers, audit domains and a structured approach to AI assurance. Exact obligations depend on the relevant requirements and risk classification.
One service built on Cygeniq’s Runtime AI Trust platform. Test and Govern form the core; Protect is available where runtime control and incident support are required.
Red-team LLMs, RAG applications and AI agents; document reproducible findings; verify that corrective actions address the underlying issue.
Maintain an AI inventory, risk classification, impact assessments, control mapping and evidence records aligned to applicable audit requirements.
Add runtime safeguards, monitoring, agent controls and incident evidence for AI systems in production.
Automated work is reviewed by Cygeniq specialists. Agent-generated output is not an independent audit opinion.
A practical approach that separates preparation and continuous assurance from independent audit judgement.
Identify and risk-tier AI systems, review priority controls, test selected models and agents, and document gaps and the action plan.
Retest after changes, monitor agreed controls and maintain findings, evidence and remediation records throughout the engagement.
An eligible CERT-In empanelled auditor examines the agreed scope and issues the audit opinion.
Risk Quotient is a CERT-In empanelled organization and our fully owned subsidiary. Connect with us to discuss your organization’s independent AI audit requirements.
The final deliverables follow the agreed engagement scope. Typical outputs include:
AI inventory with documented risk tiers and named system owners
AI risk and impact assessments and a draft Statement of Applicability
Reproducible AI red-team and control-test reports
Findings register, corrective-action plan and audit evidence pack
Relevant frameworks may include the CERT-In AI Security Auditing Guidelines v1.0, ISO/IEC 42001 and applicable RBI, SEBI, IRDAI and DPDP requirements. Scope and applicability are confirmed for each engagement.
Credit, claims, collections and customer GenAI usually land High or Critical.
Independent audits for High and Critical systems; Board oversight for Critical.
The same tiering applies to citizen services and public systems.
Audited as providers and as suppliers; customers will ask for your evidence.
Simpler paths for Low and Moderate only, with no exemption for High or Critical.
Our 90-day AI Audit Readiness Pilot focuses on three to five priority AI systems. We agree the inventory and tiering scope upfront, assess and test the selected systems, and deliver a prioritised readiness roadmap.
Yes, through Risk Quotient, our CERT-In empanelled subsidiary. Risk Quotient audits with a separate team, discloses the relationship, and never audits a control Cygeniq designed or operates; those go to another empanelled auditor. Prefer your existing auditor? They get the same re-performable evidence.
Named Cygeniq experts approve every tier, finding and evidence pack, and every agent run is recorded so your auditor can repeat it. AI output is never treated as evidence on its own.
Yes. Purchased, open-source and API-based AI are all in scope, and a vendor refusing access is never treated as proof of conformance. We assess vendors, build AI bills of materials and test the vendor model through your own interfaces.
Good, we integrate. Auditors ask for one evidence trail across all nine CERT-In domains; we fill the gaps between point tools and connect them into that trail.
The first cycle takes 13 weeks. Larger estates run in waves of the same shape, with the whole estate discovered and tiered in the first wave.
You choose SaaS, hybrid or on-premises deployment, and the data-residency model is agreed before any data is connected.
Tell us what AI you run and what you need to prepare for. We will help define a sensible starting scope for your readiness review.
AI Trust Infrastructure for secure, governed and accountable enterprise AI
© 2026 Cygeniq Inc. All Rights Reserved.