Story ·

An AI-assisted hacker took 5,200 records from an Indian university: what the report says

On 21 October 2026, Palo Alto Networks’ Unit 42 published a report on an attacker who used AI tools to attack Indian organisations. In one case the attacker took a database of more than 5,200 records from an Indian public university. Here is what the report confirms, and what it does not say.

By Threat Atlas India · Published 23 October 2026 · Based on Unit 42’s report. No university has been named.

The short version

  • Confirmed by Unit 42: an intrusion at an Indian public university, where the attacker took a MongoDB database of more than 5,200 records, including personal information on students. It also held faculty records.
  • How, as far as the report says: the attacker had an administrator account with a weak password. The report does not say how they got it.
  • The AI part: the attacker ran an AI agent framework called Hermes Agent with DeepSeek V4, and used the Cursor coding tool, with safety controls switched off in both. They relied on AI-written scripts and reports. Unit 42 says this suggests they did not have the skills such work would once have needed.
  • Not confirmed: a second Indian university’s website was attacked, but the report only says a tool flagged one login as valid. It does not say the login was used.
  • Not known: which universities, when it happened, and whether the people in the records have been told.

Confirmed versus only flagged

The report describes two Indian universities. Only one is a confirmed breach.

A login was flagged 1 login At the first university, the attacker’s tool found 12 subdomains and 13 user accounts, then tried 117 user names against 28 common passwords. It “flagged one WordPress login as valid”. The report’s caption calls it a “candidate credential”. It does not say it was used.
Confirmed intrusion 5,200+ At a second public university, the attacker took a MongoDB database of student records (diploma, undergraduate, postgraduate) and faculty records with personal information, using an admin account with a weak password.

Unit 42 does not name either university or say when the attacks happened. Threat Atlas India does not guess at either.

What the attacker did

  1. AI-written reports. The attacker’s notes are laid out like an AI’s output, with an executive summary, severity ratings and an endpoint list. Unit 42 says they used basic tools.
  2. Brute force. A generated script tried 117 user names with 28 passwords against a WordPress login. One login was flagged valid. No further use is reported.
  3. The database. The attacker took a MongoDB database. The tool’s log lists an administrator account with a weak password, student and faculty records, and exposed cloud load-balancing, database and storage accounts.
  4. Other targets. The attacker also looked at an Indian shop on a popular e-commerce platform and pulled a storefront API token, three admin email addresses, two phone numbers and payment merchant IDs. The report says targets were “primarily in India” but lists no others.
  5. A phone-fraud scheme. The attacker used an AI bot as an adviser for a fraud plan based on a US company. There were no call logs, so Unit 42 reads it as planning only.

What AI changed, and what it did not

What it changed

Unit 42 says AI let one person run several kinds of attack, including credential theft, fraud planning, Wi-Fi cracking and crypto targeting, that “would traditionally require distinct skill sets”. It says DeepSeek models are cheap, easy to get and “readily jailbroken with a simple system-prompt override”.

What it did not

The weak admin password is an old problem. Nothing in the report says AI found a new flaw. The report also does not say which AI model wrote the scripts used at the university. The attacker had other models set up in Cursor, including from other companies.

“Low-skilled” is Unit 42’s inference from the AI-generated scripts. It is not a fact about the person.

What to do

  1. Check admin accounts: make sure every administrator account on a college website, learning system or student portal has a strong, unique password and multi-factor sign-in.
  2. Close open databases: confirm MongoDB and other databases cannot be reached from the internet, and that admin consoles are not public.
  3. Hide the user list: on WordPress, stop the site from listing author names through its public API, and limit login attempts.
  4. Look for leaked secrets: scan public pages and logs for .env files, signed links, cloud keys and private keys.
  5. If you are a student or staff member: be wary of messages that quote your admission or exam details. Ask your institution in writing whether your data was involved.

What we don’t know yet

  • Which universities were affected.
  • When the intrusions happened, and for how long the attacker had access.
  • How the attacker got the administrator password.
  • Whether the first university’s login was ever used.
  • Whether the affected institutions or CERT-In have been told, and whether the students have been.

This story sticks to Unit 42’s published report and independent coverage. It will be updated if an institution or Indian authority says more.

Sources

Spotted an error? Write to [email protected]. See the method and corrections.

More stories

Breaches · 22 October 2026Amagi Media Labs: an extortion claim, what is confirmed and what is only claimedBengaluru media-technology company Amagi told the stock exchanges that a threat actor is trying to extort it over stolen network data. What Amagi confirmed, what a ransomware group lists, and what is still unknown.
Breaches · 22 October 2026Railway’s security disclosure: what to rotate, and whyRailway, a platform developers use to run their apps, says an attacker may have read customers’ stored secrets. What it says happened, what is still unknown, and what to rotate.
Breaches · 21 October 2026SBI's dark web leak: what is confirmed, and what is only claimedState Bank of India says 3,897 customer records were found in a dataset of about 3.75 lakh records on the dark web. What the bank confirmed, what the seller claimed, what the rules require, and what customers should do.
All stories →