CERT-In's AI security audit guidelines: what an auditor will check
CERT-In's AI Security Auditing Guidelines, dated November 4, 2026, sort AI systems into four risk tiers and audit them across nine domains. For AI agents, auditors will test whether an agent can exceed its permissions or bypass approval. Here is what to prepare.
Intelligent systems2 min read
CERT-In's Artificial Intelligence Security Auditing Guidelines, version 1.0, dated November 4, 2026, set out how AI systems should be audited. Each system is placed in one of four risk tiers (Low, Moderate, High or Critical), and the tier sets how often it is audited and how independently.
The four tiers decide the audit
Moderate-risk systems get annual independent audits by CERT-In-empanelled auditors. High-risk systems get periodic independent audits, which may be every six months. Critical-risk systems need CERT-In-empanelled audits and continuous monitoring.
The nine domains
- AI governance and accountability
- Data lifecycle, quality and privacy
- Model risk management
- AI ethics, bias and fairness
- Explainability and transparency
- Security, adversarial resilience and incident reporting
- Continuous monitoring, drift management and validation
- Human oversight and decision accountability
- Third-party and AI supply-chain assurance
What auditors will test on AI agents
For agents, the guidelines ask auditors to test whether an agent can exceed its authorized permissions, use tools without approval, bypass approval mechanisms or misuse credentials, and whether emergency-stop mechanisms work. These are design questions. An agent that runs under a broad service account, with approval checked only in its prompt, is unlikely to satisfy those tests.
We build AI agents with human review, so a person can approve an action before it happens where the work needs it.
Three oversight models
The guidelines name three: Human-in-the-Loop, Human-on-the-Loop and Human-over-the-Loop. Choosing between them per system, by what a wrong decision costs, is a design decision to record before the audit.
Findings come with deadlines
Moderate findings need a remediation plan within 30 business days and implementation within 60 days. Low findings must be fixed within three months. A re-audit is required when two or more Critical findings arise within three months.
What to prepare now
Start with an inventory of AI systems and assign each a provisional tier. For each one, collect what the nine domains will ask for: the data sources and purpose, validation results, monitoring, oversight model, incident procedure and vendor terms. The guidelines are aligned with ISO/IEC 42001, which helps an organization already working toward that standard.
Questions
- What are CERT-In's AI Security Auditing Guidelines?
- A framework published by CERT-In, dated November 4, 2026, for auditing AI systems, using four risk tiers and nine audit domains.
- How often must an AI system be audited under the CERT-In guidelines?
- Moderate-risk systems annually by CERT-In-empanelled auditors; high-risk systems periodically, possibly every six months; critical-risk systems with empanelled audits and continuous monitoring.
- What do the CERT-In guidelines say about AI agents?
- Auditors should test whether an agent can exceed its permissions, use tools without approval, bypass approval mechanisms or misuse credentials, and whether emergency stops work.
Sources
- Artificial Intelligence (AI) Security Auditing Guidelines, v1.0, CERT-In, November 4, 2026
- Lowdown: CERT-In's AI Security Auditing Guidelines set out tiered framework, MediaNama, November 6, 2026
- CERT-In calls for testing AI agents for unauthorised tool use, privilege abuse, The Hindu, November 5, 2026
More on intelligent systems
All writingOctober 2026
How do you test an LLM before it goes live in a regulated business?
Test a language model the way an auditor will question it: on your own cases, against written pass criteria, with attacks on its permissions and data, and with a record of every result. RBI's Governor has asked banks to red-team and stress-test AI before deployment and periodically after; five test sets are where that starts.
October 2026
Multilingual voice AI for Indian call centers: a build checklist
A voice AI line in Indian languages succeeds when it resolves the calls that make up most of the volume, hands the rest to a person with context, and logs every call in the core systems. A build checklist, with results from a line we built that answers in 11 languages.
October 2026
What drives the cost of an enterprise AI project?
The model is rarely the biggest cost in an enterprise AI project. Integration with core systems, data work, controls and the cost of running it after launch usually decide the budget. Here is how to estimate each before you sign.
Tell us which system the AI has to run inside
If that system is not ready for AI yet, tell us that too. An engineer will reply and say plainly whether we are the right firm for it.