On 30 July 2026, the government detailed its framework to counter AI-generated deepfakes — noting 13 "Responsible AI" projects approved for deepfake detection, faster removal of unlawful AI-generated content, a strengthened grievance-redressal framework, and cyber-awareness workshops reaching over 11 lakh people. For an NDA aspirant, deepfakes are a fast-rising technology-and-security topic — combining how AI works, the law, and the threat to national security.
The news in one frame
The essentials:
- What: the government's framework against AI deepfakes — legal safeguards, platform accountability and citizen protection.
- Action: 13 Responsible AI projects approved for deepfake detection; faster takedowns of unlawful content.
- Awareness: over 11.37 lakh people reached through 6,650 cyber-security workshops.
- Legal base: the IT Act, 2000 and the IT Rules, 2021.
What is a deepfake?
Start with the technology. A deepfake is synthetic media — a photo, audio clip, video or text — created or altered by Artificial Intelligence so convincingly that it appears genuine. The word blends "deep learning" and "fake." How it works:
- Deep learning — a branch of machine learning using artificial neural networks with many layers — is trained on large amounts of real images or voice recordings of a target person.
- A common method is a GAN (Generative Adversarial Network), where two neural networks compete: a generator creates fakes while a discriminator tries to detect them, so the fakes get progressively more realistic.
- The result: face-swapped video, cloned voices, or entirely AI-generated people who never existed.
Because the tools are now cheap and easy to use, deepfakes have moved from research labs into everyday misuse. This applied technology is exactly what the NDA general-knowledge notes cover.
Why deepfakes are dangerous
The examinable threat analysis:
- Misinformation & elections: fake speeches or videos of leaders can mislead voters and inflame communal tension — a threat to democracy.
- Fraud: voice cloning enables scams (a "relative" calling in distress) and corporate fraud (fake CEO instructions to transfer money).
- Personal harm: non-consensual explicit imagery and defamation, disproportionately targeting women.
- National security: fabricated statements by military or government officials could trigger panic, unrest or diplomatic incidents; adversaries can use them in information warfare.
- Erosion of trust: the deeper harm is the "liar's dividend" — when anything can be faked, people begin to doubt genuine evidence too.
These themes recur in the NDA daily current affairs.
India's legal and technical response
The examinable framework:
- Information Technology Act, 2000 — covers identity theft, cheating by impersonation, violation of privacy and publishing obscene/objectionable material electronically.
- IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 — require intermediaries (social-media platforms) to exercise due diligence, remove unlawful content within strict timelines, appoint Grievance Officers, and act on complaints about impersonation and morphed content — with loss of "safe harbour" protection if they fail.
- New criminal laws (BNS) — cover cheating, forgery, defamation and organised fraud in their modern form.
- Digital Personal Data Protection Act, 2023 — governs misuse of personal data (including images/voice used for training).
- Technical measures: AI-based detection tools, watermarking/labelling of AI-generated content, and research under the IndiaAI Mission (its Safe & Trusted AI pillar); CERT-In handles cyber-incident response.
The revision hook: deepfake = AI-generated synthetic media (deep learning + fake), often via GANs (generator vs discriminator); threats = misinformation/elections, voice-clone fraud, non-consensual imagery, national-security disinformation, "liar's dividend"; India's response = IT Act 2000 + IT Rules 2021 (platform due diligence, quick takedown, grievance officers, safe-harbour condition), BNS, DPDP Act 2023, IndiaAI Mission's Safe & Trusted AI pillar, CERT-In.
The AI-governance balance
Round out with the policy debate:
- The aim is a "safe and trusted" AI ecosystem — curbing harm without stifling innovation.
- Approaches worldwide: labelling/watermarking AI content, platform accountability, penalties for malicious use, and digital literacy so citizens can spot fakes.
- Detection is an arms race: as detectors improve, generators improve too — so law + technology + awareness must work together.
- Citizen defence: verify from credible sources, check for tell-tale glitches (odd blinking, mismatched lip-sync, strange lighting/audio artefacts), and report via the National Cyber Crime Reporting Portal (1930 helpline).
Why it matters
For the SSB and the bigger picture:
- Information warfare: in modern conflict, narrative and perception are battlefields — deepfakes are a weapon.
- Public trust: protecting the integrity of information protects democracy itself.
- Digital citizenship: awareness and critical thinking are now essential civic skills.
Exam relevance in one paragraph
For NDA General Awareness, retain: a deepfake is AI-generated synthetic media (image, audio, video or text) that convincingly mimics a real person, created with deep learning — often using Generative Adversarial Networks (GANs), where a generator and discriminator compete to make fakes more realistic; deepfakes threaten elections and democracy, enable voice-cloning fraud and non-consensual imagery, and can be used for national-security disinformation, while also creating the "liar's dividend" where genuine evidence is doubted; India counters them through the Information Technology Act, 2000 (identity theft, impersonation, obscene content), the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (platform due diligence, quick takedown, grievance officers, safe-harbour conditions), the new criminal laws (BNS), the Digital Personal Data Protection Act, 2023, plus AI-detection research under the IndiaAI Mission's Safe & Trusted AI pillar and CERT-In for cyber incidents; citizens can report cyber crime on the national portal/1930 helpline. For the SSB, it exemplifies technology, trust and national security.
🎯 Practice MCQs
Q1. A "deepfake" is media created using: (a) artificial intelligence (deep learning) (b) a typewriter (c) film cameras only (d) radio waves → (a) — AI/deep learning.
Q2. The word "deepfake" combines: (a) deep learning + fake (b) deep sea + fake (c) deep state + fake (d) deep web + fake → (a) — deep learning and fake.
Q3. A common technique used to generate deepfakes is: (a) GANs (Generative Adversarial Networks) (b) HTML (c) SQL (d) TCP/IP → (a) — Generative Adversarial Networks.
Q4. In a GAN, the two competing networks are the: (a) generator and discriminator (b) sender and receiver (c) client and server (d) anode and cathode → (a) — generator (creates) and discriminator (detects).
Q5. Deepfakes are primarily regulated in India under the: (a) IT Act, 2000 and IT Rules, 2021 (b) RTI Act (c) Forest Act (d) FEMA → (a) — the IT Act and IT Rules.
Q6. The IT Rules, 2021 require intermediaries to: (a) remove unlawful content within set timelines (b) publish all user data (c) ignore complaints (d) charge users → (a) — take down unlawful content and run grievance redressal.
Q7. Platforms failing due diligence under the IT Rules risk losing: (a) safe-harbour protection (b) their servers (c) internet access (d) employee visas → (a) — safe harbour (legal immunity for user content).
Q8. "Voice cloning" deepfakes are commonly used for: (a) financial fraud/scam calls (b) weather forecasts (c) farming (d) navigation → (a) — impersonation-based fraud.
Q9. The law governing personal data protection in India is the: (a) DPDP Act, 2023 (b) IPC (c) RTI Act, 2005 (d) Aadhaar Act only → (a) — the Digital Personal Data Protection Act, 2023.
Q10. India's national cyber-incident response agency is: (a) CERT-In (b) ISRO (c) SEBI (d) TRAI → (a) — the Indian Computer Emergency Response Team.
Q11. The mission promoting safe and trusted AI in India is the: (a) IndiaAI Mission (b) Digital India Mission only (c) Jal Jeevan Mission (d) Skill India → (a) — the IndiaAI Mission.
Q12. The national cyber-crime helpline number is: (a) 1930 (b) 100 (c) 108 (d) 1098 → (a) — 1930.
Q13. The "liar's dividend" refers to: (a) people doubting genuine evidence because fakes exist (b) profits from lying (c) a tax rebate (d) an award → (a) — erosion of trust in real evidence.
Q14. Deep learning uses which computational structure? (a) artificial neural networks (b) mechanical gears (c) analog dials (d) steam engines → (a) — multi-layer neural networks.
Q15. A technical safeguard against deepfakes is: (a) watermarking/labelling AI-generated content (b) deleting the internet (c) banning cameras (d) nothing → (a) — labelling/watermarking plus AI detection tools.
📋 How this gets asked (PYQ pattern)
AI and cyber issues are a fast-rising NDA set. The reliable framings are what a deepfake is (AI synthetic media, GANs), the governing law (IT Act 2000 + IT Rules 2021), CERT-In and the DPDP Act, and the security/misinformation threat. A common trap attributes deepfake regulation to the RTI Act or calls deepfakes simple photo-editing. The fresh 2026 hook is the deepfake-detection projects and takedown framework — ideal for "which technology / which law / which agency" items. We reference the pattern, not any exact past question.
Preparing for the NDA? AI, cyber security and information warfare are high-yield tech topics and strong SSB talking points on modern threats. Follow our daily NDA current affairs and train with serving-officer faculty in the upcoming Cavalier courses in Delhi.
✍️ Written by Maj Sunil Chopra — Co-founder & defence faculty at The Cavalier. Reviewed by the Cavalier Faculty Desk. The Cavalier, founded by ex-Army officers, has trained NDA/CDS/SSB aspirants since 2001 (Facebook · YouTube).
Source: PIB / Ministry of Electronics & IT, 30 July 2026. Facts cross-verified with independent sources.