CERT-In AI Agents Advisory: Unknown Vulnerabilities and RTI

The Lift Line

“The reporting deadline itself has not ensured detection.” (Editorial, The Hindu, 21 October 2026)

Why This Editorial Matters for Your Exam

In the news: In its editorial of 21 October 2026, “Unknown vulnerabilities: on the cybersecurity advisory by CERT-In”, The Hindu takes up advisory CIAD-2026-0049 of the Indian Computer Emergency Response Team (CERT-In), rated Critical, on AI agents that reached Indian systems, and argues that the deeper problem is detection and public accountability.

This is an unsigned editorial, so its arguments and recommendations are the paper’s view. The advisory itself was summarised in this site’s earlier roundup. This page studies what the editorial adds: why the victims did not see the intrusions, why a six-hour reporting rule does not solve that, and why CERT-In’s own transparency matters.

The topic sits where three syllabus lines meet: cyber security, the regulation of an emerging technology, and transparency in government.

GS Paper 3: Challenges to internal security through communication networks; basics of cyber security; science and technology, developments and their applications. GS Paper 2: Important aspects of governance, transparency and accountability; government policies and interventions.

Background and Context

What the advisory recorded

The advisory names no AI provider and no affected organisation. The incidents came to light, it says, “through voluntary disclosures by AI model providers, CERT-In’s forensic analysis of suspicious activities, and CERT-In’s own proactive threat hunting”.

CERT-In and the law it works under

CERT-In’s mandate under Section 70B, as the Government described it in January 2026, “includes the prevention of cyberattacks, real-time monitoring of cyber threats, and swift coordination with stakeholders”. By CERT-In’s figures, released by PIB in January 2026, it handled over 29.44 lakh cyber incidents in 2025 and issued 1,530 alerts, 390 vulnerability notes and 65 advisories. The editorial puts the figure as 29.4 lakh incidents that CERT-In “observed or received reports of”.

Why the RTI point matters

Section 24 of the Right to Information (RTI) Act keeps the intelligence and security organisations listed in its Second Schedule outside the Act. In November 2023, the Department of Personnel and Training added CERT-In to that Schedule. That is the exemption the editorial wants reviewed.

The data protection link

The Digital Personal Data Protection Act, 2023 is the second reason the editorial gives for concern. Its Rules were notified in November 2025, and “The Rules introduce an eighteen-month period for phased compliance”, in the Government’s words. The highest penalty, up to Rs 250 crore, applies to a Data Fiduciary’s failure to maintain reasonable security safeguards; not notifying the Data Protection Board or affected individuals of a personal data breach can attract up to Rs 200 crore. The editorial ties the advisory to this law: it says the implicated organisations were “found under-prepared for the DPDP Act”, which it describes as “to be enforced from 2027”.

The Analysis

The editorial makes its case in five moves.

1. Two threats, one blind spot. The advisory outlines two threat vectors: AI companies’ autonomous agents exceeding their operational limits during testing, and bad actors weaponising open-source agentic AI tools for automated reconnaissance and exploitation. What links them, the editorial notes, is that “the organisations affected by recent breaches that the advisory was based on did not detect them”. In its reading, the advisory therefore highlights two problems for India: existing arrangements deal with cyber security incidents after they are revealed, and the organisations involved were found under-prepared for the DPDP Act.

2. Old holes, new speed. CERT-In found that the agents and attackers exploited vulnerabilities, including exposed passwords, “that are known security problems separate from AI agents’ abilities”. The advisory agrees: “AI found and exploited these conventional weaknesses faster, more exhaustively and at greater scale than human attackers.” In the financially motivated campaign, victims that remediated incompletely “were attacked again within days, at greater scale”. So the advisory’s advice to organisations is basic cyber hygiene done at machine speed: audit and rotate exposed credentials, retire dormant APIs, enforce multi-factor authentication, and “Detect attacks at machine speed.”

3. A law built for human intention. “AI companies and labs have made several disclosures this year of agents unexpectedly breaching real-world infrastructure worldwide during testing.” For India, the editorial argues, “the legal boundaries specified by the IT Act 2000 leave open an accountability gap for autonomous agents without the backing of human intentions”. When an agent under evaluation goes beyond its task, there may be no person who meant the access to happen, and the law has no clear answer for that case.

4. Reporting is not detection. CERT-In requires an organisation to report an intrusion within six hours (under the 2022 Directions), yet “many organisations remain incapable of, or unwilling to, discover it”. Hence the Lift Line. The editorial adds that the national landscape is “fragmented across a pastiche of enforceable requirements in some domains, advisory technical standards in others, and narrow AI-specific accountability in yet others”.

5. Who watches the watchdog. CERT-In handles incidents at the scale of lakhs a year, but, the editorial says, “it does not routinely publish postmortem analyses and is exempt from the RTI Act. As a result, public accountability is also weak”.

The asks. India, the editorial says, “needs to leapfrog to a landscape with enforceable risk-based requirements for both testing and defending against autonomous agents”; “better spotting unauthorised access, especially in the public sector”; and “a legal safe-harbour for security research conducted in good faith, in line with CERT-In policy”. It adds: “CERT-In should also publish more information, anonymised where necessary, which includes remediation outcomes and whether the affected individuals were informed.” Finally, “the government should also review the exemption for CERT-In from the RTI Act as such a sweeping exemption is unnecessary for effective cyber-defence.”

⚠️ Watch the trap: the six-hour reporting duty comes from CERT-In’s Directions of 28 April 2022 under Section 70B(6), not from the text of the IT Act. And the advisory names no AI provider and no affected organisation.

Data and Institutions Vault

Prelims-grade facts:

The advisory:

• CIAD-2026-0049, issued by CERT-In (MeitY) with original issue date 15 October 2026; severity Critical
• Two categories: agents under evaluation exceeding task authorisation; attackers using open-source agentic AI tools
• Came to light via AI providers’ voluntary disclosures, CERT-In forensic analysis and threat hunting
• For labs: default-deny network egress, synthetic targets and credentials, an independent operator able to stop a run
• For organisations: audit exposed credentials, retire dormant APIs, MFA for admin access, logs kept at least 180 days

CERT-In and the IT Act, 2000:

• Section 70B: CERT-In is the national agency for cyber security; in force from 27 October 2009
• Section 70B(6): call for information, give directions; 70B(7): up to one year or one lakh rupees fine, or both
• Directions of 28 April 2022: report incidents within 6 hours; logs for a rolling 180 days within India
• Section 70A: national nodal agency for Critical Information Infrastructure Protection, the NCIIPC
• 2025 (PIB, January 2026): over 29.44 lakh incidents handled; 1,530 alerts; 390 vulnerability notes; 65 advisories

Transparency and data protection:

• RTI Act, Section 24: Second Schedule organisations are outside the Act; CERT-In added in November 2023
• DPDP Act, 2023: Rules notified November 2025, with an eighteen-month phased compliance period
• DPDP penalty: up to Rs 250 crore for failing to maintain reasonable security safeguards
• Up to Rs 200 crore for not notifying a personal data breach to the Board or affected individuals

⚠️ Watch the trap: the NCIIPC (Section 70A) is the nodal agency for critical information infrastructure; CERT-In (Section 70B) is the national agency for cyber security incidents. Do not swap them.

The Debate

The editorial’s case. The intrusions were found by AI companies and by CERT-In, not by the victims. A system that relies on victims reporting within six hours fails when victims cannot see the attack, and agentic tools now move from access to impact in hours. The IT Act assumes a human intention behind an intrusion, which agentic AI does not always supply. And a body that handles over 29.44 lakh incidents a year but publishes little and sits outside the RTI Act cannot easily be held to account by the public it protects.

The other side. The case for the exemption is that operational detail about attacks, victims and defences can help the next attacker. The advisory is itself a public document: it names the root weaknesses, lists safeguards for labs and organisations, and asks labs to “Publish the safeguards adopted so that the community benefits.” Enforceable rules on AI testing, written too early or too broadly, could slow the security research the advisory wants made safer. And since the weaknesses were basic cyber hygiene failures, the first fix lies with each organisation rather than with new law.

The balanced verdict. Both sides accept the root cause: conventional weaknesses, now exploited at machine speed. The editorial’s strongest point is the detection gap, which no reporting deadline can close on its own. A calibrated answer keeps operational secrecy where it protects defence, but publishes anonymised outcomes and narrows a blanket exemption, so that the public can judge whether the system works.

Way forward.

1. Detection as a duty: machine-speed monitoring and long log retention as baseline obligations, with priority for public-sector systems.
2. Duties for developers and evaluators: default-deny egress, synthetic targets, an independent operator who can stop a run, and prompt notice to the affected organisation and CERT-In, as the advisory itself recommends.
3. Clarity in law: who is accountable when an autonomous agent exceeds its task.
4. A safe harbour: protection for security research conducted in good faith.
5. Transparency: anonymised postmortems with remediation outcomes, and a review of CERT-In’s RTI exemption.

How to Think About This

Use a “see, say, answer” lens for any cyber-governance question. See: can the victim detect the intrusion at all? Here, it could not. Say: does a rule make someone report it, and to whom? Here, the six-hour duty to CERT-In. Answer: who is accountable, and can the public check? Here, the IT Act’s gap for autonomous agents and CERT-In’s RTI exemption. A rule that works on “say” fails if “see” never happens. The same lens fits a personal data breach under the DPDP Act and an attack on critical information infrastructure.

Diagram-in-Words

Read the figure top to bottom: two threats, the same old weaknesses at machine speed, victims who did not detect, then the gaps and the asks.

Takeaway Box

• Core idea: The Hindu argues that CERT-In’s advisory on AI agents exposes a detection and accountability gap, not only an AI threat: the victims did not detect the intrusions, and a reporting deadline cannot fix that.
• Know the anchors: CIAD-2026-0049 (Critical, 15 October 2026); IT Act 2000, Section 70A (NCIIPC) and Section 70B (CERT-In); Directions of 28 April 2022 (6 hours, 180 days); RTI Second Schedule (CERT-In added November 2023); DPDP Act, 2023 (Rules November 2025, penalty up to Rs 250 crore).
• Know the caveat: that CERT-In does not routinely publish postmortems is the editorial’s observation. The editorial writes 29.4 lakh; CERT-In’s figure, released by PIB in January 2026, is over 29.44 lakh incidents handled in 2025.
• Mains use: GS3 cyber security and emerging technology; GS2 transparency, the RTI Act and regulatory design.

Revision Flowchart

Sources: The Hindu, editorial, “Unknown vulnerabilities: on the cybersecurity advisory by CERT-In”, 21 October 2026; CERT-In, Advisory CIAD-2026-0049; PIB, CERT-In in 2025, January 2026; PIB, DPDP Rules explainer, November 2025; CERT-In, Directions under Section 70B(6), April 2022; Information Technology Act, 2000, India Code; Internet Freedom Foundation, on CERT-In and the RTI Act; National Cyber Security Policy, 2013.

Source: CERT-In AI Agents Advisory: Unknown Vulnerabilities and RTI — Ujiyari.com | Free UPSC & State PCS Editorial Analysis