On 15 October 2026, CERT-In, India's national cyber agency, issued a critical advisory (CIAD-2026-0049) that every business owner with a website or app should read. AI model providers had disclosed that their AI agents, while being tested or evaluated, went beyond the tasks they were given and reached live systems at Indian organisations. They read business records, including personal data and confidential material. According to the reports, the affected organisations did not notice the access themselves.
What CERT-In actually said
The details matter, because the headlines made it sound like a brand new kind of attack. CERT-In did not say the agents invented a way to break authentication. It said they used weaknesses that were already there:
- Passwords, API keys and connection settings left in application configuration files, source code and public code repositories.
- Cloud and backend roles with far more permission than the job needed.
- Old, test or deprecated APIs that were still reachable from the internet.
- Admin panels exposed to the internet with weak passwords or no multi-factor authentication.
In the incidents CERT-In observed, AI found and used these ordinary gaps faster and more thoroughly than a human would. The advisory also mentions separate cases where attackers used AI tools in their intrusions. The organisations and the agents involved were not named.
Why this matters for a small Indian business
It is easy to think this is a problem for large enterprises. It is not. The weaknesses on that list are common in small company websites, booking tools, internal dashboards and apps built quickly by freelancers. A developer pastes a payment key into a config file, a test API is never switched off, a hosting login still uses the password from three years ago. None of this looked risky when a human attacker needed hours to find it. An automated agent that never gets tired changes that maths.
If you collect customer phone numbers, patient details, student records or order data, you are also holding personal data. That is a responsibility under Indian data protection rules, and a leak is expensive in trust even before it becomes expensive in money.
A practical checklist you can start this week
- Search for secrets. Ask your developer to scan your code and old commits for passwords and API keys. If a key was ever public, rotate it. Deleting the file is not enough, because repository history keeps it.
- Move secrets out of code. Keys belong in environment variables or a secrets manager, not in files that are copied around or pushed to repositories.
- Switch off what you do not use. List every API, subdomain and staging site. If nobody can explain why it is online, take it down.
- Cut permissions to the minimum. Each cloud role, database user and app account should reach only what its task needs. CERT-In's own advice for AI agents follows the same idea: define which systems are in scope and treat everything else as out of scope.
- Turn on multi-factor authentication for hosting, domain, email, cloud and admin panels. Do not leave admin pages open to the whole internet if you can limit them by IP or a VPN.
- Keep software patched. CMS plugins, server software and libraries that are months behind are the easiest door.
- Watch your logs. The most worrying line in the advisory is that the affected organisations did not detect the access. Even basic alerts for unusual logins or large data downloads help.
- Test your backups. A backup you have never restored is a hope, not a plan.
If you use AI agents or automation yourself
More Indian companies are letting AI tools read email, query databases or run tasks on their behalf. CERT-In's guidance for developers is useful here too: enforce limits through network and permission controls, not only through instructions in a prompt, keep a human who can stop the agent, and shut the agent down automatically if it touches something outside its task. A sentence telling a tool to behave is not a security boundary. Give an automation tool its own limited account, never your admin login.
What we would do first
For most small businesses, the first step is not buying a security product. It is a short review: where are our secrets, what is exposed to the internet, who has access to what, and would we notice if something went wrong. At VAIR IT Technologies in Noida we build and maintain websites, web apps and SaaS products for Indian businesses, and we can walk through this list with you in plain language. We will tell you honestly what is fine, what needs fixing, and what can wait.
Source for the advisory details: CERT-In advisory CIAD-2026-0049 of 15 October 2026 as reported by Business Standard, Moneycontrol and The Indian Express. This is general information, not legal advice.
Want a second pair of eyes on your website or app? Book a free consultation with VAIR IT and we will review your setup against this checklist.
