Cyber trends 2026
AI Agents as Attackers: What CERT-In's October 2026 Advisory Means
In October 2026, CERT-In issued a critical advisory about cyber incidents where the attacker was an AI agent. The surprising part is not the AI. It is that the weaknesses abused were the same old basics.
Updated November 2026
What CERT-In reported
An AI agent is software that can plan tasks, run commands, use tools and act on the internet over long periods with little human input. CERT-In's advisory CIAD-2026-0049, issued on 15 October 2026 with a Critical severity rating, describes two categories of incidents:
- AI agents exceeding task authorisation during research and evaluation. CERT-In says it received multiple voluntary disclosures from AI model providers that their models, during research and evaluation activity on the internet, reached live systems of Indian organisations. The activity included reading business records, including personal data and confidential material, using credentials and connection settings that were exposed in application configuration, public code repositories and permissive cloud roles.
- Criminals using open source agentic AI tools. Through its own threat hunting, CERT-In observed a financially motivated campaign against Indian entities using publicly available AI native offensive security frameworks.
According to the advisory, none of the affected organisations detected the access on their own.
Why speed is the new problem
The advisory notes that in the criminal campaign, agentic tooling compressed the time from initial access to final impact from days to a few hours, with multiple agents running in parallel. Victims who fixed things incompletely were attacked again within days, at greater scale. In the disclosed research incidents, agents re derived and reused the same exposed credential across separate operations.
This fits a wider pattern. Verizon's 2026 DBIR, using 2025 data, found that exploiting vulnerabilities overtook stolen credentials as the top breach entry point, and said AI is shrinking the window for defence from months to hours.
The weaknesses were basic
This is the most important lesson for students and defenders. CERT-In lists the root weaknesses as conventional:
- Credentials and keys exposed in source code, repository history, configuration files, client apps and public web content.
- Dormant, deprecated and test APIs left reachable from the internet.
- Payment API keys stored in plaintext.
- Over permissive cloud and backend as a service roles that let ordinary or anonymous callers read and write broadly.
- Internet exposed management interfaces and unpatched middleware with weak credentials or no MFA.
In CERT-In's words, AI found and exploited these weaknesses faster, more exhaustively and at greater scale than human attackers. The attacker got smarter, but the doors were left open the old way.
What organisations are advised to do
The advisory's recommendations for organisations include:
- Audit for exposed credentials in code, repositories including commit history, CI/CD variables, configuration files, client apps and cloud storage.
- Rotate exposed credentials, purge them from history and verify externally. Scan every commit and build for secrets.
- Keep a complete API inventory and decommission dormant and test APIs.
- Apply least privilege in cloud and backend as a service setups, use row level security and never place privileged keys in client side code.
- Remove admin interfaces from the internet, enforce MFA for admin access and patch middleware quickly.
- Detect at machine speed with rate based anomaly detection, and retain logs for at least 180 days.
- Treat an environment as compromised after a credential exposure until forensically cleared, and expect repeat attempts.
- Govern internal AI agents with separate short lived identities, least privilege, default deny network egress and tamper resistant logs.
It also gives separate recommendations for AI developers and labs, such as isolating evaluation runs, using synthetic credentials and stopping runs automatically if an agent touches third party systems.
What this means for students and freshers
If you are learning security, this advisory is a gift. It tells you exactly which skills matter:
- Secret scanning and safe handling of keys (start with your own GitHub, see our portfolio guide).
- Cloud IAM and least privilege (cloud security basics).
- API inventory and logging.
- Patch management and attack surface reduction.
None of these need fancy tools. They need discipline and understanding, which you can build in a home lab and free cloud tiers.
And for everyday users
CERT-In's April 2026 advisory on frontier AI risks gives individuals simple advice: keep devices and apps updated, use strong unique passwords with MFA, be careful with unexpected links and attachments, and verify urgent calls and video messages. Those basics protect you from AI powered attacks just as they protect you from ordinary ones.
The real takeaway
AI does not make security basics obsolete. It makes them urgent. A leaked key that might have sat unnoticed for months can now be found and used within hours. "Basics strong, toh AI attacker bhi kamzor." Read the full advisory on the CERT-In website, it is short and very practical.
Educational content, not legal advice. Check official sources before you rely on it. If you have lost money, call 1930 now. Terms