Evidence-first authority page
Security Review Questions for an AI Agent Deployed in India
What a security review of an AI agent typically covers beyond incident reporting, without asserting any product's security posture.
Direct answer
A security review of an AI agent covers more ground than incident-reporting obligations alone — prompt-injection exposure, tool-call permission scope, and what a compromised agent session can actually reach are separate questions from whether an incident gets reported to CERT-In after the fact. This page lists the questions a review typically raises; it does not attest that any specific product passes them.
Last reviewed 2026-09-22
Reviewed by Febin Francis, Founder/Editor on 2026-09-22 under the methodology-only publication path: this page makes no externally-verifiable factual or comparative claim, so it is reviewed for originality and disclosed limitations rather than sourced against a primary receipt.
Questions beyond incident reporting
- What can the agent's tools actually reach — a scoped API key, or broad account access? See agent sandbox and trust boundary in this site's glossary.
- Is user-supplied content (a document, a webpage, an email) treated as untrusted input, or can it alter the agent's instructions? See prompt injection defense.
- Is there a limit on how many tool calls or how much spend an agent session can generate before a human is alerted?
- How is a security finding disclosed, and does that process meet the buyer's own obligations under CERT-In's directions.
Why this stays a question list, not a verdict
A security posture is specific to one product's current implementation and changes with each release — this site does not run its own penetration tests, so it records what a vendor's own security documentation states (where one exists) rather than asserting a security property independently.