Military AI and urgency of guardrails

Military AI and the Urgency of Guardrails

UPSC Prelims + Mains Study Note | GS-II / GS-III

1. At a Glance

• Military AI refers to the application of artificial intelligence to defence systems — from logistics and intelligence to Lethal Autonomous Weapons Systems (LAWS), which can select and engage targets without direct human control. [S1]
• Governance of military AI sits at the intersection of technology regulation, international humanitarian law (IHL), and geopolitics — making it high-value for GS-II (IR) and GS-III (Security & Technology).
• Dual-use nature of AI (civilian + military R&D pipelines overlap) makes verification of any arms-control regime exceptionally difficult. [S4]
• India's strategic ambiguity — abstaining at REAIM 2026, voting against a UN First Committee resolution on LAWS (2023) — is a recurring theme in recent international affairs. [S2][S5]

2. Why in the News

• February 19, 2026: The third REAIM Summit (Responsible AI in the Military Domain) concluded; only 35 of 85 participating countries signed the 'Pathways to Action' declaration — down from 60 signatories at the previous summit. India, the US, and China all declined to sign. [S4]
• The abstention occurred just days before the India AI Impact Summit, drawing attention to the gap between India's civilian AI ambitions and its military AI governance posture. [S4]
• December 19, 2024: UN Secretary-General addressed the Security Council specifically on AI, urging restraint in military applications. [S6]
• UN General Assembly Resolution 79/239 and subsequent Resolution 80/58 marked a shift toward implementation-level measures on AI and security. [S3]
• UNIDIR Global Conference on AI, Security and Ethics scheduled for 18–19 June 2026, Palais des Nations, Geneva — further internationalising the debate. [S3]

3. Background & Evolution

Predecessor frameworks: CCW GGE on LAWS (since 2014); US DoD Directive 3000.09 (Autonomous Weapons, 2012, updated 2023); NATO AI principles (2021).

4. Core Static Facts

Definitions & Terminology

• LAWS (Lethal Autonomous Weapons Systems): Weapons that select and engage targets autonomously without meaningful human control.
• Human-in-the-loop: Human authorises each individual engagement.
• Human-on-the-loop: Human can override but system acts autonomously in real time.
• Human-out-of-the-loop: Fully autonomous — no human intervention possible.
• Dual-use technology: AI R&D serves both civilian and military ends simultaneously, complicating export controls and treaty verification. [S4]
• REAIM: Responsible Artificial Intelligence in the Military Domain — multilateral summit process, not a treaty body.

Key Bodies / Processes

India's Formal Positions

• India is a High Contracting Party to the CCW and participates in GGE on LAWS. [S2]
• India voted against UN First Committee LAWS resolution, 2023 (148 countries voted in favour). [S5]
• India did not sign REAIM 'Pathways to Action', Feb 2026. [S4]

Key Numbers

• REAIM 2026: 85 participating countries, 35 signatories (~41%).
• REAIM 2024 (Seoul): ~60 signatories.
• UN Secretary-General's target: prohibitions/restrictions on autonomous weapons by 2026. [S6]
• UN GA Res. 79/239 and 80/58 on AI in the military domain. [S3]

5. Multi-Dimensional Analysis

Geopolitical / Strategic

• Great-power abstention (US, China, India) from REAIM 2026 signals that leading AI-capable states resist binding constraints, fearing strategic disadvantage. [S4]
• Russia-Ukraine and Gaza conflicts have accelerated deployment of AI-enabled drones, target recognition systems, and electronic warfare — demonstrating real-world urgency.
• India's abstention is partly strategic hedging: preserving freedom of action in border contexts (China, Pakistan) while engaging multilateral forums. [S2]
• A binding LAWS ban would disproportionately constrain states that lack AI capacity to develop compliant systems — a concern for India's Make in India Defence ambitions. [S8]

Legal / Constitutional (International Humanitarian Law)

• IHL principles at stake: Distinction (civilian vs. combatant), Proportionality, Precaution — LAWS risk violating all three if targeting algorithms are opaque.
• UN Secretary-General: "An algorithm must not be in full control of decisions involving killing." [S5]
• No country should use military AI that violates IHL or human rights law — UN consensus position. [S6]
• CCW GGE operates by consensus, allowing any major power to block progress indefinitely.

Ethical / Governance

• Accountability gap: If a LAWS kills civilians, who is responsible — the programmer, commander, or state? Existing war-crimes law has no clear answer.
• Meaningful human control (MHC) is the central governance concept; defining it operationally remains contested.
• India's article (Adya Madhavan, Takshashila Institution) argues India should push for a non-binding framework rooted in accountability principles aligned with its interests — between total abstention and premature hard commitments. [S4]
• Transparency of AI decision-making in targeting is near-impossible with deep learning models (black-box problem).

Scientific / Technological

• AI is dual-use: same computer vision, autonomous navigation, and NLP stack underpins consumer tech and weapons systems. Verification regimes face fundamental technical limits. [S4]
• DRDO active in autonomous systems and robotics (technology foresight domain). [S8]
• AI in warfare: ISR (Intelligence, Surveillance, Reconnaissance), cyber operations, logistics, predictive maintenance, and direct kinetic systems (loitering munitions, autonomous drones).

Administrative

• India has no dedicated Military AI policy document in the public domain as of 2026.
• MEITY governs civilian AI regulation; MoD/DRDO handles defence R&D — inter-ministerial coordination gap on dual-use AI.
• India's IndiaAI Mission (2024, ₹10,371 crore) focuses on civilian AI infrastructure; military AI governance is not addressed. [S9]

Historical

• Precedents: Anti-Personnel Mines Treaty (Ottawa, 1997), Chemical Weapons Convention (1993) — both achieved broad prohibition but not universal adherence (US, Russia not party to Ottawa).
• Pattern: Great powers resist prohibition; opt for codes of conduct — same dynamic now playing out in LAWS.
• Nuclear non-proliferation analogy: Early abstainers (India on NPT, 1968) later faced pressure but also preserved strategic autonomy.

6. Recent Developments (Last 12–18 Months)

• October 2024: UN First Committee passes resolution on LAWS; India had voted against a similar resolution in 2023. [S5]
• December 19, 2024: UN Secretary-General addresses Security Council on AI, calls for urgent governance measures and reiterated ban target by 2026. [S6]
• February 2026: REAIM Summit 3 — signatory numbers drop from ~60 to 35; India, US, China do not sign 'Pathways to Action'. [S4]
• June 2026 (current month): UNIDIR Global Conference on AI, Security and Ethics at Palais des Nations, Geneva — examining implications of AI for international peace and security; UNGA resolutions 79/239 and 80/58 cited as milestones shifting from principles to implementation. [S3]
• UN News (June 2025): Growing pressure internationally to regulate 'killer robots' amid proliferation in conflict zones. [S9]
• India AI Impact Summit held shortly after REAIM 3, highlighting domestic–international governance dissonance. [S4]

7. Prelims Hooks

1. REAIM stands for Responsible Artificial Intelligence in the Military Domain. [S4]
2. The first REAIM Summit was held in The Hague in February 2023. [S7]
3. The second REAIM Summit was held in Seoul in September 2024. [S7]
4. At REAIM 2026 (third summit), only 35 of 85 participating countries signed the 'Pathways to Action' declaration. [S4]
5. India, the United States, and China were among countries that did not sign REAIM 2026. [S4]
6. India voted against the UN First Committee resolution on LAWS in 2023 (148 countries voted in favour). [S5]
7. The primary UN forum on LAWS is the Group of Governmental Experts (GGE) under the Convention on Certain Conventional Weapons (CCW). [S1]
8. UN Secretary-General's stated target: binding prohibitions/restrictions on autonomous weapons by 2026. [S6]
9. UN General Assembly Resolution 79/239 is a key resolution on AI in the military domain. [S3]
10. India's nodal ministry for international disarmament and LAWS negotiations is the Ministry of External Affairs (Disarmament & International Security Affairs division). [S2]
11. DRDO is India's implementing agency for autonomous systems and robotics R&D. [S8]
12. The concept of "meaningful human control (MHC)" is the central principle in LAWS governance debates. [S5]
13. UNIDIR (UN Institute for Disarmament Research) organises the Global Conference on AI, Security and Ethics (June 2026, Geneva). [S3]
14. AI is classified as dual-use technology — having both civilian and military applications — making treaty verification inherently difficult. [S4]
15. The Takshashila Institution (think tank) published the article arguing India should pursue a non-binding accountability framework for military AI. [S4]

8. Mains Relevance

GS Paper Mapping

Plausible Mains Question Stems

1. "The declining number of signatories at successive REAIM Summits reflects fundamental tensions in governing military AI. Critically analyse India's strategic position and suggest a coherent policy framework." (GS-II / GS-III, 250 words)
2. "Lethal Autonomous Weapons Systems pose unprecedented challenges to International Humanitarian Law. Examine the adequacy of existing multilateral mechanisms and India's role in shaping governance norms." (GS-II, 250 words)
3. "The dual-use nature of artificial intelligence renders traditional arms-control verification approaches obsolete. Discuss with reference to global efforts to regulate military AI." (GS-III, 150 words)

9. Related Topics to Study Next

1. Convention on Certain Conventional Weapons (CCW) — direct parent framework for LAWS negotiations; India's position within it.
2. India's AI Policy & IndiaAI Mission — domestic context; why civilian AI governance and military AI governance are decoupled.
3. Nuclear Non-Proliferation Treaty (NPT) and India's outsider status — historical analogy for India abstaining from major arms-control regimes.
4. International Humanitarian Law (IHL) — Geneva Conventions — legal bedrock against which LAWS compliance is tested.
5. Cyber Warfare and India's National Cyber Security Policy — overlapping domain; AI used in offensive cyber operations.
6. DRDO and Indigenisation of Defence Technology — Make in India Defence; how autonomous systems R&D fits.
7. UN Secretary-General's Common Agenda / Pact for the Future — broader UN reform framework within which AI governance sits.
8. Dual-Use Technology Export Controls (Wassenaar Arrangement) — how AI hardware/software is currently (partially) regulated in trade.

10. Common Errors / Trap Areas

1. REAIM ≠ a treaty or binding agreement — it is a summit process producing political declarations, not legally binding instruments. Aspirants often conflate it with formal arms-control treaties.
2. Confusing REAIM summit numbers: Summit 1 = The Hague (Feb 2023); Summit 2 = Seoul (Sep 2024); Summit 3 = Feb 2026. The article refers to "the third global summit."
3. India's vote vs. India's abstention: India voted against (not merely abstained from) the UN First Committee LAWS resolution in 2023 — a stronger negative signal than abstention.
4. MEITY ≠ nodal body for military AI — MEITY governs civilian AI; military AI governance sits with MoD/MEA, not MEITY. Confusing these is a common trap.
5. LAWS ≠ all military AI: LAWS refers specifically to weapons that select and engage targets autonomously. Military AI also covers logistics, ISR, predictive maintenance — which face different and less contentious governance questions.

11. Sources

• [S1] Summit on Responsible Artificial Intelligence in the Military Domain — https://en.wikipedia.org/wiki/Summit_on_Responsible_Artificial_Intelligence_in_the_Military_Domain — (Tier 3/reference)
• [S2] Disarmament and International Security Affairs Overview — Ministry of External Affairs, Government of India — https://www.mea.gov.in/05-disarmament-and-international-security-affairs-overview — (Tier 1)
• [S3] UNIDIR Global Conference on AI, Security and Ethics 2026 — https://indico.un.org/event/1023183/overview — (Tier 2, UN)
• [S4] "Military AI and urgency of guardrails" — The Hindu, 19 February 2026 (article excerpt, primary source supplied) — https://www.thehindu.com/todays-paper/2026-02-19/th_international/articleGSKFJVG8C-13571893.ece — (Tier 4)
• [S5] First Committee Approves New Resolution on Lethal Autonomous Weapons — UN Press, 2023 — https://press.un.org/en/2023/gadis3731.doc.htm — (Tier 2, UN)
• [S6] Secretary-General's Remarks to Security Council on Artificial Intelligence, December 2024 — https://www.un.org/sg/en/content/sg/statements/2024-12-19/secretary-generals-remarks-the-security-council-artificial-intelligence-bilingual-delivered — (Tier 2, UN)
• [S7] Minister of National Defence Remarks at REAIM Summit 2024 — Government of Canada — https://www.canada.ca/en/department-national-defence/news/2024/09/minister-of-national-defence-remarks-at-the-responsible-ai-in-the-military-domain-reaim-summit-2024.html — (Tier 3)
• [S8] Autonomous Systems and Robotics — DRDO, Ministry of Defence, Government of India — https://drdo.gov.in/drdo/en/offerings/technology-foresight/autonomous-systems-and-robotics — (Tier 1)
• [S9] As AI evolves, pressure mounts to regulate 'killer robots' — UN News, June 2025 — https://news.un.org/en/story/2025/06/1163891 — (Tier 2, UN)