Why in news?
A CERT-In advisory flagged rogue autonomous AI agents and AI-powered attacks, exposing weak breach detection in Indian organisations. The affected organisations did not detect the breaches themselves. These came to light only through AI companies’ disclosures and CERT-In’s own operations.
UPSC Relevance
Prelims: CERT-In, IT Act 2000, DPDP Act 2023, agentic AI, cyber attack terms.
Mains: GS3 – Challenges to internal security through communication networks, basics of cyber security; Science and Technology (developments in AI).
What is Cybersecurity?
Cybersecurity is the practice of protecting computers, networks, programs and data from unauthorised access, attack, damage or theft. It rests on three basic principles, known as the CIA triad:
- Confidentiality: Only authorised people can access data.
- Integrity: Data is accurate and not tampered with.
- Availability: Systems and data are accessible when needed.
India’s Cybersecurity Framework at a Glance
| Law / Institution | Key provisions |
|---|---|
| IT Act, 2000 | Main cyber law; Section 43 (damage to computer systems), Section 66 (computer offences), Section 66F (cyber terrorism), Section 69A (blocking) |
| CERT-In (Section 70B) | National nodal agency for incident response under MeitY; 2022 Directions require incidents to be reported within 6 hours and logs kept for 180 days; placed in the Second Schedule of the RTI Act in 2023 |
| DPDP Act, 2023 | Data fiduciaries must take reasonable security safeguards and report breaches; penalty up to Rs 250 crore; full enforcement from 2027 under DPDP Rules, 2025 |
| I4C (MHA) | Indian Cyber Crime Coordination Centre; runs helpline 1930 and the cybercrime.gov.in portal |
| National Cyber Security Coordinator | Coordinates cyber security at the national level under the National Security Council Secretariat |
| National Cyber Security Policy, 2013 | Only national policy so far; a new national cyber security strategy is still awaited |
Recent incidents in India: AIIMS Delhi ransomware attack (2022), malware detected at Kudankulam nuclear plant’s administrative network (2019), and reports linking the 2020 Mumbai power outage to a foreign state-linked group.
How AI is Changing Cyber Threats?
Artificial intelligence works both ways. It strengthens defence, but it also gives attackers speed, scale and automation. The biggest change is agentic AI, i.e. AI systems that can plan and carry out multi-step tasks on their own, such as scanning networks, writing code and using tools, with little human supervision.
| AI-related threat | How it works |
|---|---|
| Rogue autonomous agents | AI agents under testing act beyond their authorised limits and access real systems, as flagged by CERT-In |
| Automated reconnaissance and hacking | Open-source agents scan thousands of systems, find weak passwords or unpatched software, and exploit them at machine speed. In 2025, an AI company disclosed a largely AI-run espionage campaign by a state-linked group |
| Deepfakes and voice cloning | Fake audio or video used for fraud; in 2024 a Hong Kong firm’s employee was tricked into transferring about USD 25 million after a deepfake video call |
| AI-powered phishing | Personalised, error-free scam messages produced at huge scale in local languages |
| Polymorphic malware | Malicious code that keeps changing its form to escape antivirus detection |
| Prompt injection | Hidden instructions in a webpage or document hijack an AI assistant into leaking data or taking harmful actions |
| Data poisoning | Corrupting training data so an AI model behaves wrongly or carries a hidden backdoor |
| Shadow AI | Employees pasting confidential data into unapproved AI tools, leading to leaks |
Why AI threats are harder to handle
- Lower entry barrier: Even unskilled actors can now run sophisticated attacks.
- Speed gap: Attacks happen in minutes while human defenders take days or weeks to respond.
- Attribution problem: When an autonomous agent causes harm without human intent, it is unclear who is liable – the developer, deployer or user.
- Dual use: The same tools used by security researchers to find flaws can be used by criminals.
Gaps in India’s Approach
- Detection over reporting: Many organisations, especially in the public sector, lack continuous monitoring. Reporting rules cannot work without detection.
- Fragmented rules: Binding requirements in some sectors (RBI, SEBI, IRDAI), advisory standards in others, and narrow AI-specific rules elsewhere.
- No AI-specific liability: The IT Act does not clearly cover harm from autonomous systems. India’s AI Governance Guidelines (2025) are principle-based, not binding.
- Low preparedness for DPDP Act: Many entities do not yet have reasonable safeguards, despite steep penalties from 2027.
- Transparency deficit: CERT-In’s RTI exemption and lack of published post-mortems weaken public accountability and shared learning.
- Skills shortage: India faces a large gap in trained cybersecurity professionals.
Strategies to Deal With the Threat
- Risk-based enforceable norms: Mandatory safety requirements for testing and deploying autonomous agents, with stricter rules for high-risk uses.
- Clear liability: Update the IT Act (or the proposed Digital India Act) to fix responsibility of developers and deployers for AI agent actions.
- Safe harbour for researchers: Legal protection for good-faith vulnerability disclosure, in line with CERT-In’s own policy.
- National cyber security strategy: Replace the 2013 policy with a strategy that covers AI, cloud and critical infrastructure.
- Shift to proactive detection: 24×7 Security Operations Centres, threat hunting and AI-based anomaly detection, especially in government departments.
- Zero trust architecture: ‘Never trust, always verify’ – every user and device is checked continuously.
- Publish incident learnings: CERT-In should release anonymised post-mortems, remediation outcomes and whether affected people were informed.
- Review RTI exemption: A blanket exemption is not needed for effective cyber defence; sensitive details can be protected case by case.
- Public-private partnership: Real-time threat information sharing with AI companies, banks and telecom firms.
- Skilling: Large-scale training of cyber professionals and awareness drives like Cyber Surakshit Bharat.
- International cooperation: Work with partners and global forums on AI safety norms and cybercrime, since attacks cross borders.
Practice MCQ
Q1. With reference to the Indian Computer Emergency Response Team (CERT-In), consider the following statements:
1. It is the national nodal agency for cyber incident response under Section 70B of the Information Technology Act, 2000.
2. It functions under the Ministry of Home Affairs.
3. It has been placed in the Second Schedule of the Right to Information Act, 2005.
How many of the statements given above are correct?
(a) Only one
(b) Only two
(c) All three
(d) None
Answer: (b)
Explanation: Statement 2 is incorrect. CERT-In works under the Ministry of Electronics and Information Technology (MeitY). The Indian Cyber Crime Coordination Centre (I4C) is under MHA.
Q2. Consider the following pairs:
| Term | Description |
|---|---|
| 1. Prompt injection | Hidden instructions that make an AI system act against its intended purpose |
| 2. Data poisoning | Manipulating training data to corrupt an AI model’s behaviour |
| 3. Zero trust architecture | Granting full access to all users once they are inside the network |
How many of the pairs given above are correctly matched?
(a) Only one
(b) Only two
(c) All three
(d) None
Answer: (b)
Explanation: Pair 3 is incorrect. Zero trust means no user or device is trusted by default, even inside the network; every access request is verified.
Mains Practice Question
Q. The rise of agentic artificial intelligence has made cyber attacks faster, cheaper and harder to attribute. Examine the gaps in India’s cybersecurity framework in this context and suggest measures to strengthen it. (15 marks, 250 words)
Tell Google you want more of this.
Add Anantam IAS as a preferred sourceOne tap, and this site shows up more often in your own Top Stories, AI Overviews and AI Mode. Remove it any time.
