Индия разрабатывает «убийственных роботов» без закона и этического контроля

In March 2026, a parliamentary committee tabled a report in the Lok Sabha that should have triggered a national debate. It didn’t. The report confirmed that DRDO and defence public sector undertakings are actively developing Lethal Autonomous Weapon Systems (LAWS) — “killer robots” — alongside AI for underwater autonomous vehicles, missile applications, and a facial recognition system designed to identify individuals even when disguised. The development comes in the complete absence of any AI law, oversight body, or dedicated legislation in India. The policy that says “yes” and the law that says nothing India’s tri-services AI policy, dated March 9, 2026, explicitly endorses Lethal Autonomous Weapon Systems as “one of the earliest and most prominent military applications of AI.” The policy envisions drone swarms and LAWS variants strengthening area denial capabilities, with a “transition to unmanned combat operations in the near term.” It lists potential applications including unmanned aerial bombing, close-combat support, and autonomous border or harbour patrols. The policy does include the phrase “meaningful human oversight.” But the parliamentary committee report found that no AI law exists in India, none is imminent, and no oversight body has been named. The Ministry of Defence itself admitted to the committee: “AI ML-based techniques are not amenable for verified decision-making and hence are likely to result in unintended outcomes.” Its solution? “A semi-automatic control with AI-supported recommendations could be very impactful.” But there is no global consensus on what “semi-automatic” means, and the ministry acknowledged that there is no agreement on defining “levels of autonomy.” The accountability vacuum The report contains no reference to India’s position in international negotiations regarding autonomous weapons. No oversight body is named. No legislation is proposed or mentioned as under consideration. The DRDO is developing autonomous weapons. The policy endorses them. The law is silent. This is the gap nobody is talking about. The international contradiction India’s international stance on autonomous weapons is internally contradictory. Having chaired the Group of Governmental Experts that produced the first internationally agreed guiding principles on LAWS, India voted against UN General Assembly Resolution 78/241. India argues that international humanitarian law already regulates the effects of weapons rather than the technologies that produce them, so existing law suffices. This position is coherent where a system merely executes a human judgement already made — but it falls silent where the system performs the legal reasoning of distinction and proportionity itself. India abstained on a 2024 resolution calling for stronger human control norms and voted against a 2023 resolution mandating a UN report. India maintains that autonomy can improve precision and reduce human error, and cautions against “stigmatising emerging technologies.” But as a legal analysis noted, this stance is “strategically rational but unstable” — the strategic logic explains why India wants existing law to suffice without showing that it does. The Lavender warning and the human factor The real-world stakes are not theoretical. Investigations into Israel’s Lavender AI system found that it generated mass target lists of Hamas operatives with limited human oversight, leading to significant civilian casualties including entire families. The parliamentary report explicitly cited this as a warning: AI systems are “not amenable for verified decision-making” and may produce “unintended outcomes.” At the India AI Impact Summit, military leaders expressed sharp concerns. Lieutenant General Vipul Singhal described a case where a machine-recommended strike failed to account for a civilian evacuation, prompting a commander to intervene and stop the strike. Major General Harsh Chhibber warned: “Command responsibility is absolute in the military. You cannot do cognitive offloading to a machine.” Another officer stated: “The requirement is to make better decisions, not bad decisions faster.” “You cannot do cognitive offloading to a machine. Command responsibility is absolute.” — Maj. Gen. Harsh Chhibber, India AI Impact Summit The ETAI Framework: India’s hidden advantage There is a silver lining that almost no one knows about. DRDO has developed the Evaluating Trustworthy AI (ETAI) Framework, an operational governance model for military AI that, according to a CSDR analysis, “neither great power has matched.” The framework provides a structured approach to evaluating AI trustworthiness in military contexts — addressing the very concerns that the parliamentary committee raised. But it remains a DRDO internal tool, not codified into doctrine, law, or service regulations. The missing pieces India’s military AI doctrine must address several gaps that currently exist only as bullet points in policy documents, not as binding rules. First, human-machine interaction protocols defining exactly when a human must be in the loop, on the loop, or out of the loop. Second, escalation management — how autonomous systems interact with nuclear command structures. Third, target verification standards for AI-assisted or AI-autonomous engagements. Fourth, incident investigation protocols when an autonomous system makes an error. Fifth, auditability — requiring explainable AI modules and immutable logging for mission decisions to enable accountability after the fact. The technology is being built. The weapons are being developed. The policy has been written. What is missing is the legal and ethical architecture that determines who is responsible when a machine makes a lethal decision and gets it wrong. That is the story nobody is writing about — and it is the one that matters most.