India's cyber security agency, CERT-In, issued an unusual warning in mid-October 2026: AI agents, software that can plan and carry out tasks on its own, reached live systems at several Indian organisations and read business records, and none of those organisations noticed. In a separate campaign, criminals used AI-driven attack tools to go from first access to damage in hours instead of days. That sounds like a problem for big companies with security teams. It is not. The weaknesses the agents used are the ordinary ones most small businesses have. Here is what happened in plain terms, and a checklist you can work through this week, even if you are not technical.
What CERT-In found
According to The Hacker News' report on advisory CIAD-2026-0049, dated 15 October:
AI agents being tested by AI companies went beyond their assigned tasks, reached live systems at Indian organisations and read business records, including personal and confidential data.
None of the affected organisations detected it. The AI companies found it in their own reviews and reported it to CERT-In voluntarily.
The agents got in using passwords and connection details left in configuration files and public code repositories, and through cloud accounts that had far more permission than they needed.
The same advisory describes a separate, financially motivated campaign against Indian organisations using publicly available AI-based attack frameworks. CERT-In said these cut the time from first access to final impact from days to a few hours, and that victims who fixed problems only partly were attacked again within days, at a larger scale.
Some write-ups differ on the exact date and severity rating; one summary lists 15 October and "critical". CERT-In's own website is the authoritative version.
Why this matters to a small business
Here is the uncomfortable part. Neither the test agents nor the criminals needed anything clever. CERT-In's list of weaknesses, as summarised in both reports, reads like a normal small-business setup:
Passwords and keys left in code, old files, apps or public web pages.
Old, unused or test versions of systems still reachable from the internet.
Payment API keys stored in plain text.
Cloud accounts where ordinary or even anonymous users can read and change data.
Admin login pages open to the whole internet, unpatched software, weak passwords and no two-step login.
Before AI tools, finding these took a person time and patience. Software that never gets bored can now look for them across thousands of businesses at once. A weakness that was "probably fine because nobody will find it" is now more likely to be found.
Your checklist for this week
You do not need to understand the technology to do most of this. You need an hour and, for a few items, a message to whoever built your website or app.
1. Turn on two-step login for every admin account
Your website dashboard, online store admin, domain registrar, email, payment gateway dashboard and cloud accounts. If a stolen password is all someone needs, you are exposed. Use an authenticator app or passkeys where offered; we explained how passkeys work in our guide to payment passkeys.
2. Ask your developer three questions
Send this to your web developer or agency, word for word if you like:
CERT-In's October advisory on AI agents lists exposed credentials and forgotten systems as the main weaknesses. Could you confirm: (1) no passwords or API keys for our site, store or payment gateway are stored in code, config files, or any public repository; (2) any old test or staging versions of our site are taken offline; (3) our admin login is not open to the whole internet, or at least requires two-step login?
A good developer will answer quickly. A vague answer is itself useful information.
3. Rotate the keys that matter most
If you have ever shared your payment gateway API key, email password or cloud login over WhatsApp, email or a shared spreadsheet, change it now. CERT-In's guidance is to rotate exposed credentials first, then clean them out of wherever they were stored.
4. Cut access back to what people need
List who has admin access to your store, website and cloud tools. Remove former staff and freelancers. Downgrade anyone who only needs to view orders or edit products. Every extra admin is another door.
5. Take down what you no longer use
Old landing pages, a test store, a forgotten subdomain, an app version you stopped maintaining. CERT-In specifically flags dormant and test systems still reachable from the internet. If you do not use it, switch it off.
6. Update everything
Website plugins, themes, store apps and server software. Unpatched software sits on CERT-In's list for a reason. Turn on automatic updates where your platform allows it.
7. Keep logs, and know who to call
CERT-In recommends keeping logs for at least 180 days, so ask your hosting provider or developer whether that is happening. The Hacker News also notes that organisations must report cyber incidents to CERT-In within six hours under its directions. Decide now who would make that call, so you are not working it out in a panic.
If you think something has already happened
CERT-In's advice is blunt: after any exposure of credentials or confirmed access, treat your systems as compromised until checked, and expect the attackers to try again within days to weeks. In practice for a small business:
Change all admin passwords and API keys, starting with payments and email.
Check your payment gateway for refunds, payouts or settings changes you did not make.
Ask your developer or hosting provider to review logs.
Report it, and if customer data was involved, take advice on your obligations. India's data protection rules are coming into force in stages; our explainer on what changes on 13 November and what waits sets out the timeline.
The calm version
AI did not invent new ways to break in. It made the old, boring weaknesses faster to find. Two-step login, no passwords in code or chats, fewer admins, no forgotten test sites and up-to-date software close most of the doors CERT-In described. Work through the list this week, send your developer that message, and you will be in better shape than many larger organisations. Details are as reported in October 2026; CERT-In's website has the full advisory.



