UPSC Mains GS4 Question Paper 2026 with PDF Solution

What was actually delegated

The professor did not use a tool to save labour. He delegated judgement — the scholarly assessment of whether a thesis meets the standard of a doctorate — to a system that cannot hold an opinion, cannot be examined on it, and cannot be held responsible for it. The later “modifications” do not repair this, because editing an output is not the same as forming a judgement and then expressing it.

Why accountability cannot be transferred

• Accountability attaches to a role, not to a task. The university appointed a person, not a process. Signing the report is a declaration that the assessment is his. See our note on accountability and responsibility for the distinction between answerability and mere task completion.
• The candidate cannot appeal to an algorithm. A doctoral evaluation carries the right to a reasoned assessment by a qualified peer. If the reasoning was never held by a human mind, that right is hollow.
• Answerability requires an author. If the report is later challenged, no one can explain why a particular chapter was judged inadequate, because nobody formed that view.

The integrity breach is separate from the accuracy question

Suppose the AI report is excellent. The wrong survives, and this is the point most answers miss. Integrity is the alignment between what one represents and what one has done. The professor represents the report as his professional assessment; it is not. The breach is one of honesty and non-disclosure, and it exists whether or not the content is sound.

• Concealment is the aggravating fact. Undisclosed use is what converts assistance into deception.
• Fiduciary duty. The candidate, the university and the discipline all rely on the assessment being what it claims to be.
• Institutional harm. If undisclosed AI evaluation becomes normal, the doctorate stops certifying anything.

Paucity of time is a reason, not a justification

Workload pressure is real and it explains the choice. It does not justify it, because the honest options were open: seek an extension, decline the assignment, or disclose the method. Choosing silence over any of these is what makes it a lapse rather than a hard call.

What would have made it defensible

• Disclosure to the university and the candidate that AI assisted the drafting.
• Substantive human judgement — the professor reads the thesis, forms the assessment, and uses the tool only to organise or express it.
• An institutional policy on permissible AI use in evaluation, which most Indian universities still lack. Our note on AI governance in India covers the emerging framework.

Conclusion

The tool is not the problem; the silence is. Using AI to draft an assessment one has genuinely formed is efficiency. Using it to manufacture an assessment one has not formed, and presenting it as one’s own, is a failure of both accountability and integrity — and no amount of subsequent editing converts the second into the first.

The conflict, stated precisely

This is not a clash between development and tribal welfare. It is a clash between an aggregate gain that is countable — investment, jobs, revenue — and a concentrated loss that is not — a community’s habitat, livelihood and social world. The asymmetry is what makes it hard: the gains are visible in a file, the losses are visible only if someone goes and looks.

The ethical challenges I would face

• Utilitarian arithmetic against rights. The project may raise aggregate welfare while extinguishing the way of life of a specific group. Rights exist precisely to stop such trade-offs from being automatic.
• Consent that is formal but not free. A gram sabha resolution obtained under pressure, or with incomplete information, satisfies the file and not the principle. Our note on the PESA Act 1996 covers where that consent sits in law.
• Conflicting loyalties. Political executive, corporate proponent, affected community and my own conscience will pull in different directions, and each has a legitimate claim on my attention.
• Compensation that cannot restore. Money can replace a house. It cannot replace a forest that supplies food, medicine, fuel and identity. Treating cash as full settlement is a category error.
• Inducement and capture. Pressure will rarely be crude. It will arrive as an appeal to the district’s development ranking.
• Intergenerational effects. The children of the displaced inherit the loss without having been party to any consent.

What is not discretionary

Several duties here are legal, not ethical choices, and an officer who treats them as negotiable has already failed.

• Forest Rights Act, 2006 — recognition and vesting of individual and community forest rights must precede diversion; see our note on the Forest Rights Act.
• PESA gram sabha consultation in Scheduled Areas.
• Free, prior and informed consent where the group is a particularly vulnerable tribal group, whose habitat rights are protected.
• Environmental clearance conditions and rehabilitation entitlements under the LARR Act, 2013.

How I would hold the tension

• Make the invisible countable. Commission a proper social impact assessment so the loss enters the record with the same authority as the projected revenue.
• Insist on process integrity — a genuine gram sabha, in the local language, with the project’s real footprint disclosed.
• Search for the third option. Realignment, reduced footprint or benefit-sharing often exist and are not looked for because nobody is asked to look.
• Put dissent on the file. If overruled, a written, reasoned note preserves both the record and my own integrity.

Conclusion

The district officer is not the arbiter of whether the project proceeds. He is the guarantor that the community’s rights are not quietly traded away in the process — and that guarantee is discharged through the record he creates, not through the outcome he prefers.

The answer, and the reason

No. The District Magistrate should not administer an untested supplement to the tribe, however promising the research and however severe the deformity. The reason is not that the intervention might fail. It is that a DM using administrative authority to enrol a vulnerable population into an untried treatment is conducting an experiment without calling it one — and therefore without any of the protections an experiment requires.

The four principles, applied

• Autonomy. Consent must be free and informed. A community that is endangered, unwell and dependent on the administration cannot meaningfully refuse a DM. The power differential makes real consent very hard to obtain and easy to fake.
• Non-maleficence. Primum non nocere is the operative principle where the evidence is thin. Mineral supplementation is not harmless — excess selenium, fluoride or iron causes serious toxicity, and dosing in an undernourished population is unpredictable.
• Beneficence. The intent to relieve suffering is genuine, but beneficence is measured by expected benefit on available evidence, not by hope.
• Justice. A group is being asked to bear the risk of an untested intervention because it is poor, remote and easy to reach administratively. That is the historical pattern of research abuse, not a departure from it.

Why vulnerability raises the bar

The instinct that “these people have nothing to lose” inverts the principle. In research ethics, vulnerability triggers additional safeguards, not fewer: independent review, enhanced consent, and a positive justification for why this group and no other. Our note on medical ethics for the administrator develops this.

The administrative ethics dimension

• Role boundary. A DM is not a clinician or an ethics committee, and should not become one by default because the health system is absent.
• Coercion by office. Even a sincere request from a DM carries the weight of the state.
• Accountability if it goes wrong. The researcher publishes; the community bears the harm; the DM signs nothing. Diffuse responsibility is itself a warning sign.

What the DM should actually do

• Route it through the proper channel — ICMR ethical guidelines, an institutional ethics committee, and a registered clinical trial with DCGI approval where required.
• Use administrative power for what it is good at: getting a specialist medical team to the area, establishing the actual cause through proper investigation, and fixing water, nutrition and access in the meantime.
• Treat the deformity as a public health emergency, which permits urgent action on established interventions without requiring an untested one.
• Document and escalate so the state health department owns the problem rather than a single officer improvising.

Conclusion

Compassion and urgency argue for doing something; they do not argue for doing this. The ethical course is not inaction but correct action — bringing the health system to the tribe rather than bringing an experiment to it.

Reject the zero-sum framing first

The question is usually posed as a trade-off in which every unit of rights surrendered buys a unit of security. That framing is wrong, and starting from it produces a weak answer. Rights violations frequently reduce security: custodial torture produces unreliable intelligence, mass surveillance buries signal in noise, and communities that experience the state as an adversary stop supplying the information that actually prevents attacks. Security and rights are complements far more often than they are substitutes.

Where the tension is nevertheless real

• Speed against process. Preventive action must sometimes precede the evidence that would satisfy a court.
• Secrecy against transparency. Sources and methods cannot be disclosed, which weakens the ordinary machinery of accountability.
• Group suspicion against individual guilt. Profiling is operationally tempting and morally corrosive.
• Emergency against permanence. Extraordinary powers are taken in a crisis and rarely surrendered afterwards. This is the deepest problem.

The tests that discipline the trade-off

The Supreme Court’s proportionality test in K. S. Puttaswamy (2017) is the working instrument, and a strong answer names its limbs:

• Legality — the restriction must rest on a law, not on executive convenience.
• Legitimate aim — national security qualifies, but it must be the actual aim, not a label.
• Necessity — no less restrictive measure would achieve it.
• Proportionality stricto sensu — the harm to the right must not exceed the security gain.
• Procedural safeguards — review, time limits and a route to challenge.

Institutional safeguards that make it real

• Judicial oversight of detention, interception and asset seizure.
• Sunset clauses so emergency powers expire unless actively renewed.
• Independent review — the NHRC, and parliamentary scrutiny of the intelligence apparatus, which India still lacks.
• Custodial protections — the D. K. Basu guidelines, videography and time limits.
• Data minimisation under the Digital Personal Data Protection Act, 2023, though its wide state exemptions are a live criticism.

The Indian record, honestly

India has both strong constitutional doctrine and persistent practice problems: low conviction rates under UAPA alongside long pre-trial detention, and the Supreme Court’s own observation that bail is the rule even in terror cases. The gap is not in the principle but in its enforcement.

Conclusion

Balance is not a midpoint between two goods. It is a discipline: every security measure must survive the proportionality test, carry a sunset, and answer to someone outside the agency that wants it. A state that meets those conditions is usually more secure, not less.

What Ambedkar actually said

On 25 November 1949, in his closing address to the Constituent Assembly, Ambedkar urged that India abandon “the bloody methods of revolution” and also “the method of civil disobedience, non-cooperation and satyagraha”, calling them “the Grammar of Anarchy”. A weak answer stops there. A strong answer notices the clause he attached: these methods were to be abandoned because constitutional methods were now open. Under the Raj there had been no constitutional route, which is why he had himself used agitation; once there was one, extra-constitutional means lost their justification.

The condition is the whole argument

Ambedkar’s position is therefore conditional, not absolute. It holds while constitutional channels are genuinely available and genuinely responsive. The modern dilemma arises exactly where that condition weakens — where legislation is passed without deliberation, where courts are slow, where a grievance has no electoral constituency. Our note on Ambedkar’s philosophy develops the constitutional-morality argument underneath it.

The tests that separate legitimate protest from anarchy

Rawls’s account of civil disobedience supplies workable criteria, and they map well onto Indian conditions:

• Exhaustion of lawful means — representation, litigation and electoral remedy tried first.
• Public and open — done in the open, not covertly, because it is an appeal to the community’s sense of justice.
• Non-violent — violence converts an appeal into coercion.
• Aimed at a substantial injustice, not at any policy one dislikes.
• Willingness to accept the legal penalty, which is what demonstrates fidelity to the legal order as a whole.
• Proportionality — the disruption imposed on uninvolved citizens must bear some relation to the wrong protested.

How a modern democracy should navigate it

• Distinguish dissent from disorder. The state’s duty is to protect the first while policing the second, and conflating them is the standard failure.
• Keep the constitutional channel credible. Pre-legislative consultation, functioning committees and timely courts are the real prophylactic against street politics.
• Respond to the grievance, not only to the disruption. Clearing a road settles nothing if the claim is unanswered.
• Restraint in the use of force, and proportionate rather than punitive prosecution.
• Protect the uninvolved. A protest that blocks access to hospitals or livelihoods loses the moral standing it claims.

The honest tension

Ambedkar’s warning and the Gandhian tradition genuinely pull against each other, and pretending otherwise weakens an answer. Our note on Gandhi and Ambedkar sets out the disagreement. The resolution most defensible today is procedural: civil disobedience is ethically available as a last resort, under the tests above, and is illegitimate as a first resort in a system that offers real alternatives.

Conclusion

Ambedkar was warning against making agitation the ordinary grammar of politics, not against conscience itself. A democracy navigates this by keeping constitutional remedies fast and real — because the surest way to produce anarchy is to leave a genuine grievance with nowhere lawful to go.

What trusteeship claims

Gandhi’s doctrine holds that the wealthy do not own their surplus; they hold it in trust for society and are answerable for how it is used. Ownership becomes stewardship, and the test of the steward is not intention but the condition of the beneficiary. Our note on Gandhi’s philosophy covers the wider framework.

The translation to public office

The doctrine transfers almost exactly, because an officer’s position is already a trust in law and not property. Authority, discretion, information and public funds are all held for beneficiaries who cannot supervise their use.

• Discretion is trust property. A transfer, a licence, a tender or a posting is not the officer’s to give as a favour. Treating discretion as personal capital is the root of most administrative corruption.
• The beneficiary is the whole public, weighted towards those least able to press their claim.
• Accountability runs to the beneficiary, not only upwards to the hierarchy.

Concrete applications

• Antyodaya as the decision rule. Gandhi’s talisman — recall the face of the poorest person you have seen and ask whether your step is of any use to him — converts trusteeship into an actual test at the moment of decision.
• Transparency as reporting to the beneficiary. Proactive disclosure under the RTI Act, published beneficiary lists and social audit are trusteeship in institutional form; MGNREGA social audits are the clearest Indian example.
• Refusing benefit from the office. No use of position, information or hospitality for personal advantage — trusteeship implies non-possession applied to power.
• Stewardship of resources. Public money spent as a trustee would spend it, and natural resources held for future generations rather than allocated for present convenience.
• Succession thinking. A trustee leaves the institution stronger than he found it — systems, records and trained subordinates rather than personal indispensability.
• Means and ends together. A fair outcome reached by an unfair process breaches the trust, because the beneficiary is owed both.

The honest criticism

Trusteeship was attacked by Ambedkar and by socialists as naive: it relies on the voluntary conscience of the powerful and supplies no remedy when conscience fails. That criticism lands against trusteeship as economic policy. It lands much less against trusteeship in administration, where the trust is legally enforceable — through service conduct rules, the Prevention of Corruption Act, RTI, audit and judicial review. In governance, Gandhi’s moral idea has an institutional backstop that his economic version never had.

Conclusion

For an administrator, trusteeship is not a sentiment but a rule of construction: every power is read as held for someone else. An officer who applies it treats discretion as borrowed, transparency as reporting, and the condition of the weakest citizen as the measure of whether the trust has been discharged.

Where the line comes from, and who says it

The line is from Henry IV, Part 1, Act V, Scene 2, and the speaker is Hotspur — a rebel, addressing his troops before Shrewsbury, justifying an armed revolt against his king. That provenance is the most useful fact in the question. The claim that just intent makes arms fair is placed in the mouth of someone using it to license rebellion, and he loses. Shakespeare is not endorsing the maxim; he is showing how easily it is reached for.

The defensible core

Intent is not irrelevant. Law and ethics both grade conduct by it: mens rea distinguishes murder from accident, and the doctrine of double effect distinguishes a foreseen harm from an intended one. An officer who acts in good faith on the record before him is judged differently from one who acts to favour a friend, even where the outcome is identical.

Why intent alone will not do

• Everyone believes their intent is just. Intent is self-certified and unfalsifiable, which makes it the weakest possible check on power.
• It licenses the ends-justify-means slide. If just purpose sanctifies the instrument, encounter killings, custodial coercion and fabricated files all become available to a sincere officer.
• Consequences fall on others. The displaced, the wrongly detained and the excluded bear the cost of a good intention that went wrong.
• Means shape ends. Gandhi’s insistence that means and ends are convertible terms is the sharpest counter to Hotspur: a just order cannot be built by unjust instruments, because the instruments become the order.

The three-part test

A defensible decision must satisfy all three, not one:

• Intent — is the purpose legitimate and public-regarding?
• Means — are they lawful, proportionate and the least restrictive available?
• Consequences — are the foreseeable effects, including on those not party to the decision, acceptable?

What this means in governance

• Recorded reasons. Because intent cannot be inspected, the file must carry the reasoning that stands in for it.
• Due process is not an obstacle to good intent; it is the only way a claim of good intent can ever be tested.
• “For the greater good” needs a stated cost. An officer invoking it should be able to say who bears the loss and why that is acceptable.
• Guard against noble-cause corruption — the well-documented pattern where officers bend procedure believing the cause justifies it. Our note on the framework for resolving ethical dilemmas covers the working method.

Conclusion

Hotspur’s maxim is half true and dangerous in the half that is false. Just intent is necessary to make an act defensible; it is nowhere near sufficient. Governance is the discipline of requiring the other two — lawful means and acceptable consequences — from people who are already certain their intentions are good.

The two concepts, sharply distinguished

• Beneficence is the duty to act for the patient’s good. It is about the content of the decision.
• Paternalism is overriding or bypassing a competent person’s own choice for that person’s good. It is about who decides.

They are commonly confused because paternalism is usually motivated by beneficence. The distinction is that beneficence asks “is this good for her?” while paternalism asks “may I decide it for her?” — and a doctor can be right on the first and wrong on the second.

What the doctor actually did

He had genuine clinical grounds: long-term medication does carry serious adverse effects, and recommending surgery is sound practice. But he did not persuade the patient. He persuaded the family, and used the family to move the patient. That routing is the decisive act, and it converts beneficence into hard paternalism.

• Soft paternalism overrides a choice that is not fully voluntary or informed — a patient who has misunderstood the diagnosis, say. It is often defensible.
• Hard paternalism overrides the informed choice of a competent adult. It is presumptively wrong, and this is that.

Why the family route is the core wrong

Nothing in the facts suggests the patient lacked capacity. She had a considered preference for medication. Bringing family pressure to bear does not correct a defect in her reasoning; it applies leverage to her circumstances. In the Indian setting this is especially significant for female patients, where family decision-making can substitute for a woman’s own choice without anyone intending coercion. The autonomy breach is structural, not merely procedural.

Is reluctant consent valid consent?

Signed, yes. Valid, doubtfully. Consent must be free as well as informed, and the word “reluctantly” records the absence of the first. Indian law has moved firmly this way: in Samira Kohli v. Prabha Manchanda (2008) the Supreme Court held that consent must be real and specific to the procedure, and rejected the substituted judgement of relatives for a competent adult patient. See our note on medical ethics.

What beneficence without paternalism looks like

• Address the patient directly, with the risks of both options set out in terms she can weigh.
• Disclose uncertainty honestly, including the risks of surgery, not only of medication.
• Offer time and a second opinion; urgency was not pleaded here.
• Involve family only with her permission, and as support rather than as a lever.
• Accept a refusal. A competent adult may choose the medically inferior option, and honouring that is the test of whether autonomy was ever real.

Conclusion

The doctor’s intention was beneficent and his clinical judgement may well have been correct. He nevertheless acted wrongly, because he treated the patient’s consent as an obstacle to be managed rather than a decision to be respected. Beneficence justifies making the strongest possible case to the patient; it never justifies going around her.

The distinction, fixed

Drucker’s formulation is the standard one: efficiency is doing things right; effectiveness is doing the right things. Efficiency is a ratio of output to input. Effectiveness is a judgement about whether the output was worth producing at all. Productivity, properly understood, is the product of the two — and a high-efficiency, low-effectiveness system produces a great deal of nothing very useful.

Why the second failure is worse

An inefficient organisation doing the right thing wastes resources. An efficient organisation doing the wrong thing accelerates towards the wrong destination, and its own metrics will report success while it does so. This is the more dangerous condition because it is invisible from inside.

• Toilets built but unused for want of water — efficient construction, ineffective sanitation.
• Classrooms with high enrolment and low learning — the ASER findings are precisely an effectiveness failure sitting behind efficient enrolment numbers.
• Grievance portals that close complaints fast without resolving them, because disposal rate is what is measured.

Why administration is biased towards efficiency

• Efficiency is measurable and attributable; effectiveness is diffuse, slow and often visible only after a transfer.
• Goodhart’s law — once a measure becomes a target it stops being a good measure. Targets on files disposed, funds utilised or cases closed reliably produce exactly those numbers.
• Rule-orientation. Following procedure correctly is defensible; asking whether the procedure achieves anything is not required of anyone.

Striking the balance

• Set the purpose before the target. State the outcome the scheme exists to produce, then choose indicators that track it — learning outcomes rather than enrolment, water at the tap rather than pipeline kilometres.
• Use outcome budgeting, which India has adopted precisely to force this question at the point of allocation.
• Pair every efficiency metric with an effectiveness metric so neither can be gamed alone.
• Build feedback from the beneficiary — social audit, citizen charters and independent evaluation tell you whether the right thing happened, which internal reporting will not.
• Periodic purpose review. Ask whether the activity should continue at all; sunset the ones that should not.
• Sequence correctly. Effectiveness is the prior question, efficiency the subsequent one. Deciding what to do comes before deciding how well to do it.

The honest caveat

Effectiveness is not an excuse for indiscipline. An officer who neglects timelines and process while claiming to pursue outcomes usually delivers neither. The balance is not a compromise between the two but a hierarchy: choose the right objective, then pursue it efficiently. Our note on good governance covers the institutional machinery for this.

Conclusion

Productivity in public administration is not files moved per day. It is the amount of the intended public outcome produced per unit of public resource — and any system that measures only the first will optimise itself away from the second.

The choice, correctly stated

The dilemma is not rules versus compassion. It is a conflict between two obligations of the same administrator: to enforce standards that exist to protect children, and to secure the right to education of children who have no alternative school within reach. Closing a non-compliant school in a remote area does not move those children to a compliant one; it moves them out of education.

Why literal enforcement fails

• It defeats the purpose of the rule. Infrastructure and teacher norms under the RTE Act exist to make schooling effective. Enforcing them into closure produces no schooling at all.
• The burden is regressive. The affected children are the poorest and most remote — precisely those the Act was written for.
• Irreversibility. A child out of school at ten is often out permanently, and girls disproportionately so. The harm does not wait for compliance.

Why non-enforcement also fails

• Standards protect children too. An unsafe building, an untrained teacher or an absent toilet are not paperwork problems.
• Selective non-enforcement corrodes the system and becomes a rent-seeking opportunity.
• It entrenches a two-tier entitlement in which remote children are permanently owed less.

The graded approach

The defensible course is neither closure nor blindness, but a triage that distinguishes what cannot be compromised from what can be sequenced.

• Separate safety from paperwork. Structural safety, drinking water, toilets, and protection from abuse are non-negotiable and must be fixed immediately. Playground area, boundary walls and record formats can be phased.
• Time-bound compliance plans with named milestones, in writing, rather than either a closure notice or silence.
• Fix the state’s own default first. Many deficiencies are the government’s failure — unfilled teacher posts, undelivered grants. Penalising a school for the state’s omission is unjust.
• Use recognition conditionally, which the RTE framework allows, rather than as a binary.
• Alternatives before closure. If a school must close, arrange transport, residential facilities or a neighbouring school first; the sequence matters more than the decision.
• Document the deviation. Record the reasoning and the timeline so this is a considered administrative judgement and not a favour.

The ethical reasoning underneath

The child is the beneficiary of both the rule and the exception, so the child’s interest is the tie-breaker. Rigid rule-following here would be what Weber warned of as bureaucratic ritualism — the means displacing the end. Equally, unstructured discretion invites capture. The resolution is principled discretion: deviate openly, for stated reasons, against a deadline, with the record available for scrutiny.

Conclusion

An administrator’s duty is to the purpose of the rule, not only its text. Where the two diverge, the honest course is to protect what the rule protects — the child — while creating a time-bound, documented path back to full compliance.

Concede the description first

The claim is largely accurate as description. States sanction some invasions and not others, court some autocracies and isolate others, and speak of human rights selectively. Any answer that denies this is unserious. The interesting question is not whether it happens but why, and whether it is defensible.

Three positions on why

• Realism. The international system is anarchic — no authority above states — so survival is the first duty and a state that indulges morality at the cost of security fails its own citizens. Morgenthau argued that the state has no right to sacrifice its people for an abstract principle. Our note on key concepts in international relations covers this.
• Liberal institutionalism. Cooperation, treaties and reputation change the calculus, so ethical behaviour is often also strategically rational over time.
• Constructivism. Interests are not fixed and given; they are shaped by identity and norms. What a state counts as its interest today was constructed by earlier normative argument — the abolition of the slave trade and the taboo on chemical weapons being the clearest cases.

Why the dichotomy is overstated

• Reputation is a strategic asset. A state seen as unreliable pays for it in every subsequent negotiation.
• Ethical positions can be interest-serving. India’s opposition to apartheid cost little and bought lasting standing in Africa.
• Norms constrain even the powerful — not absolutely, but the need to justify shows the constraint is real. States lie about compliance rather than announcing indifference.
• Global public goods — pandemic response, climate, sea lanes — cannot be secured by narrow self-interest, and the interest itself becomes collective.

India’s record, honestly

India illustrates both sides and pretending otherwise weakens an answer. On the ethical side: leadership of the anti-apartheid movement, non-alignment as a refusal to be instrumentalised, Vaccine Maitri, and consistent championing of the Global South’s development claims. On the interest side: continued oil purchases from Russia after 2022 despite Western pressure, abstentions at the UN on Ukraine, and a calibrated position on Myanmar’s junta driven by border and connectivity concerns. Our note on the foundations of India’s foreign policy covers the doctrinal basis.

What disciplines the trade-off

• Consistency. Not perfect consistency, which is impossible, but the avoidance of nakedly opposite positions on identical facts.
• Transparency about the reason. Saying “this serves our energy security” is more honest, and more defensible, than dressing it as principle.
• Red lines. Some acts — genocide, chemical weapons use, targeting of civilians — should not be traded for advantage, and a state gains by being known to hold them.
• Proportionality. The interest served should bear some relation to the ethical cost imposed.

Conclusion

Ethics in international relations is neither sovereign nor irrelevant. It functions as a constraint that raises the cost of certain choices rather than forbidding them — and a state that is honest about when it is paying that cost, and holds a small number of genuine red lines, is both more credible and, over time, more secure.

What is actually wrong here

The question invites an answer about empathy against rules. That framing should be resisted, because the decisive fact is elsewhere: the teacher tutors the Personal Assistant’s daughter. The DM did not act on a general policy about children with learning disabilities. He acted on the request of a subordinate who personally benefits from the outcome. That is a conflict of interest, and it is present whatever the DM felt.

The three distinct wrongs

• Private benefit through public power. A transfer order is public authority. Using it so that one official’s daughter keeps her tutor is the diversion of a public instrument to a private end.
• Improper channel. A PA has no locus in transfer decisions. Acting on his request, rather than on a representation from the school, the parent body or the education department, corrupts the decision process regardless of the merits.
• Arbitrariness under Article 14. Other children in the district with learning disabilities, and other teachers with equal claims, received no such consideration. A benefit available only to those with access to the DM’s office is by definition arbitrary.

Why the empathy framing misleads

Empathy is the capacity to understand another’s situation and let it inform judgement. It is a genuine administrative virtue, and our note on empathy and compassion sets out why. But empathy has to be impartial to be a virtue at all. Feeling for the person in front of you while being blind to the identical claims of people not in the room is not empathy; it is proximity bias. The child here is genuinely deserving. So are the children whose teachers were transferred without anyone speaking for them.

The rule-compliance side, fairly stated

Rules on transfers exist to prevent exactly this — favouritism, tenure manipulation and the capture of postings by the well-connected. A rigid transfer policy that ignores every human circumstance is also defective; the answer is not that rules must never bend.

What genuine empathy would have looked like

• Convert the individual favour into a general rule. If continuity of a trained teacher matters for children with disabilities, issue a district policy — under the Rights of Persons with Disabilities Act, 2016 and the RTE inclusive-education mandate — that applies to every such child.
• Address the real need. The daughter needs assessment, an individualised education plan and a trained special educator, not one teacher’s private tuition. Solving it through a transfer freeze treats a symptom.
• Recuse from the specific case, or route it through the education officer, once a subordinate’s interest is known.
• Record the reasons if a deviation is genuinely warranted, so it can be tested.
• Separate the private tuition question — teachers taking private tuition is itself often barred by service rules, which the arrangement quietly ignores.

Conclusion

Empathy and compliance are not really in conflict here, because this was not an act of empathy. It was a favour that felt like one. The administrative test is simple and worth stating: would this decision have been taken, on these facts, for a family with no connection to the DM’s office? If not, it was not compassion.

The claim and where it comes from

This is Peter Singer’s argument, made in “Famine, Affluence and Morality” (1972). Its structure is short and hard to escape:

• Suffering and death from lack of food, shelter and medical care are bad.
• If it is in our power to prevent something bad from happening, without sacrificing anything of comparable moral importance, we ought morally to do it.
• Absolute poverty is preventable at a cost that is not comparably important to an affluent person.
• Therefore we ought to prevent it — and failing to do so is wrong, not merely less than admirable.

The drowning child

Singer’s illustration does the work. If you pass a shallow pond where a child is drowning, you must wade in even though it ruins your clothes. Nobody thinks this is optional generosity, and nobody thinks the cost of the clothes is a defence. Singer’s challenge is that distance and the presence of other bystanders are the only differences between that case and a preventable death abroad — and neither is a morally relevant difference. Our note on Peter Singer develops this.

The distinction the question asks for

• Duty (obligation) is owed. It is claimable by the beneficiary, its breach warrants blame, and it is not discharged by inclination. Kant’s perfect duties are of this kind.
• Charity (supererogation) is above and beyond. It is praiseworthy to do, not blameworthy to omit, and the giver chooses recipient, amount and timing.

The moral consequence of reclassification is the whole point. Under charity, the affluent person who gives nothing is simply not generous. Under duty, the same person has wronged someone. The beneficiary moves from grateful recipient to holder of a claim.

The case for treating it as duty

• Capacity generates obligation. We already accept this in rescue duties, in the duty to report a crime, and in the special obligations of doctors and public servants.
• Omission can be culpable. The law itself recognises this where a duty to act exists; morality recognises it more broadly.
• Charity leaves the poor dependent on mood. Rights-based welfare — the entitlement approach, the NFSA, MGNREGA — exists precisely because dignity requires claims rather than favours.
• Indian traditions converge. Dana in the Gita is enjoined as duty when given to a fit recipient without expectation, and zakat is obligatory rather than voluntary — both treat giving as owed, not optional.

Objections, and honest replies

• Demandingness. Taken strictly, the principle requires giving until further giving costs something comparable, which is close to impoverishing oneself. Singer’s moderate version — the qualifier in this question, “without significantly affecting your financial stability” — is the defensible one.
• Causation. I did not cause the famine. True, but the drowning-child case shows that the duty to rescue does not depend on having caused the peril.
• Effectiveness. Aid is sometimes wasted. This argues for giving well, not for not giving.
• Special obligations. Family and community come first. Granted — but this limits the scope of the duty, it does not dissolve it.

Conclusion

The claim is justified in its moderate form. Where the cost to me is trivial and the benefit to another is survival, the asymmetry is so extreme that calling my response “charity” understates what is at stake. The strongest version of the argument is not that we must give everything, but that the language of charity lets us treat as optional something we already know is owed.

Stakeholders and their claims

• Lata — the patient, a competent adult, whose bodily autonomy and reproductive capacity were at stake.
• Dr Mansi — clinician, owing beneficence, non-maleficence and honesty.
• Sujatha — sister-in-law, present and willing, but with no legal authority over a competent adult.
• The hospital and the profession — whose consent protocols were tested.
• Future patients — for whom this precedent decides whether their consent means anything.

The ethical issues involved

• Scope of consent. Lata consented to a diagnostic laparoscopy. A hysterectomy is a different procedure with permanent, life-altering consequences. Consent to diagnosis is not consent to treatment.
• Invalid surrogate consent. Sujatha’s written agreement has emotional weight and no legal force. For a competent adult, no relative can consent on her behalf; temporary unconsciousness under anaesthesia does not create incapacity in the relevant sense, because the patient could have been asked before or after.
• Autonomy against beneficence. Dr Mansi optimised for Lata’s physical welfare — sparing her a second surgery and its risks — at the cost of her right to decide about her own body.
• The emergency exception does not apply. It covers situations where delay threatens life or limb. A suspected malignancy awaiting biopsy is serious but not immediately life-threatening; the operation could have been closed and the decision taken with Lata awake.
• Reproductive autonomy and gender. Removing a woman’s uterus without her word engages her dignity and reproductive rights, and reflects a pattern in which women’s medical decisions are routinely routed through family.
• Honesty and disclosure — informing Lata only the next day compounded the breach.

The precedent that decides it

These facts are nearly identical to Samira Kohli v. Prabha Manchanda (2008), where a patient consented to a laparoscopy and the surgeon performed a hysterectomy after obtaining the mother’s consent. The Supreme Court held the surgery unlawful: consent must be real, specific to the procedure, and given by the patient herself; a relative’s consent for a competent adult is no consent at all. That Lata’s outcome may have been clinically better is irrelevant to the legality. See our note on medical ethics.

Assessing Dr Mansi’s moral conduct

Her conduct is best described as well-intentioned but ethically wrong, and the answer should say both.

• In her favour — no self-interest, a genuine clinical rationale, she did not act secretly, she consulted the person present, and she took written consent rather than none.
• Against her — she chose the convenient consent over the valid one, treated an anaesthetised patient as decision-incapable when she was merely temporarily unconscious, and let her own clinical certainty substitute for the patient’s judgement about her own life.

The police view that good intention settles the matter is mistaken, and Lata’s sense of betrayal is a proportionate response to a real wrong, not an overreaction.

What she should have done

• Pre-operative anticipatory consent — discuss, before surgery, what should happen if a tumour is found, and record Lata’s instruction. This is the single step that would have prevented everything.
• Close and take a biopsy, then present the finding to Lata and let her choose, accepting the burden of a second surgery as her decision to make.
• Document the reasoning and inform her immediately on waking, not the next day.

Conclusion

A competent adult’s consent is not an administrative formality to be satisfied by the nearest available signature. Dr Mansi acted to protect Lata’s health and, in doing so, took from her the thing the law and ethics protect most closely — the right to decide what is done to her body. The remedy lies in anticipatory consent protocols, not in asking patients to be grateful for outcomes they never chose.

The dilemma

Ravi has a system that is working by its own metrics — public order has visibly improved — and a credible allegation that the improvement rests on discriminatory foundations. The trap is that the evidence of success and the evidence of injustice are the same data.

The ethical issues, including bias

• Feedback loops. The model is trained on historical arrest data. If a neighbourhood was over-policed before, it generated more recorded crime, so the model directs more patrols there, which generates more arrests, which confirms the model. This is a ratchet, not a discovery — the system measures police attention, not crime.
• Proxy discrimination. Even without ethnicity as an input, postcode, income and migration status act as proxies. The claim that the algorithm is neutral because it is blind to identity is false.
• Presumption of innocence inverted. Preventive detention and checkposts based on statistical association punish people for correlations, not conduct.
• Opacity and due process. Residents cannot see what is recorded against them, cannot contest it, and cannot correct it. A person marked by an unseen file has no remedy.
• Consent and privacy. Biometric capture of everyone in a crowd is mass surveillance of the innocent, engaging the proportionality test from Puttaswamy (2017). Our note on surveillance covers the wider problem.
• Accountability gap. If the system is wrong, who answers — the vendor, the officer who acted, or nobody?
• Erosion of police legitimacy. Order maintained by surveillance of a community produces compliance without consent, and destroys the cooperation on which real policing depends.

Options before Ravi

• Reject the memorandum and continue. Merit: results, deterrence, no operational disruption. Demerit: entrenches bias, invites litigation, and treats a rights complaint as an obstacle.
• Suspend the system entirely. Merit: immediately stops the harm and signals good faith. Demerit: loses genuine investigative value, may raise crime, and over-corrects on an unproven allegation.
• Continue but add safeguards without independent scrutiny. Merit: preserves capability. Demerit: self-audit by the beneficiary of the system is not credible to the very community that complained.
• Suspend the automated enforcement triggers, keep the system for investigation only, and commission an independent audit while engaging the community.

The recommended course, and why

The fourth option optimises ethical compliance while preserving legitimate capability. Concretely:

• Immediately stop AI-driven preventive detention and blanket checkposts. No liberty should be restricted on an algorithmic score alone; require independent human grounds, recorded, for each action.
• Retain the tool for post-crime investigation — matching a suspect against a library after an offence is a different act from marking a neighbourhood before one.
• Commission an independent bias audit — technical experts plus civil society, examining training data, error rates disaggregated by community, and false-positive burden.
• Open the data to the data subject. A notice-and-correction mechanism so residents can see and contest entries.
• Publish the policy — purpose, retention period, access controls, and the human-review requirement.
• Engage the delegation as partners, not adversaries: a standing community-police committee with visibility of aggregate outcomes.
• Address the underlying conditions. Gang violence in a low-income immigrant neighbourhood has causes that patrolling does not touch.

Conclusion

Ravi should treat the memorandum as intelligence rather than as opposition. A predictive system that cannot be audited, contested or explained does not become acceptable because it works; and a police force that maintains order by making one community permanently suspect has traded long-term legitimacy for short-term statistics. Our note on police reforms sets out the institutional context.

The dilemma

Seema faces a network, not a few corrupt individuals. Touts operate through subordinate staff, honest employees are frightened into inaction, and the resulting backlog harms the public independently of the bribery. Both available instincts fail: a frontal assault risks institutional paralysis and union backlash, while tolerance entrenches the racket and prolongs public harassment.

The ethical issues

• Corruption as extortion. Citizens are paying for a service they are entitled to receive. The wrong is not only illegal gain but the conversion of an entitlement into a purchase.
• Regressive incidence. Large bribes screen out the poorest applicants entirely, so the racket allocates rights by wealth.
• Abuse of discretion. Selective, non-chronological approvals are the visible signature of discretion being sold. Chronology is not a bureaucratic nicety; it is the main defence against favouritism.
• Fear as an instrument of capture. Honest officials are disabled — a serious harm to them and to the office.
• Public harm from the backlog — economic progress halted and citizens harassed, which is a cost of inaction that must be weighed.
• Duty to act. Knowledge creates obligation; a Collector aware of an organised racket who does nothing becomes complicit.
• Due process for the accused. Suspicion is not proof, and mass punishment would be its own injustice.

Options

• A. Immediate mass suspension and police complaint. Merits: decisive, signals zero tolerance. Demerits: acts on suspicion without evidence, invites union action and stay orders, empties the office of trained staff, and stalls the very backlog she must clear.
• B. Ignore or quietly transfer the suspects. Merits: no disruption. Demerits: moves the problem, rewards the guilty, demoralises honest staff, and abandons the public.
• C. Systemic reform only — digitise and time-bound the process without pursuing individuals. Merits: removes the opportunity, no confrontation. Demerits: impunity for past extortion, and networks adapt to new systems if nobody is held to account.
• D. Sequenced approach — reform the process first to stop the bleeding, build evidence quietly against the ringleaders, then act with proof while protecting honest staff.

The recommended option: D, and its sequence

Option D is recommended because it treats the two harms — the corruption and the backlog — as needing different instruments, and because it is the only option that does not trade one for the other.

• Close the discretion. Move CLU applications online with mandatory chronological processing, a public dashboard of pending files with dates, and system-enforced deadlines under the state Right to Services Act. A tout’s product is queue-jumping; publish the queue and the product disappears.
• Clear the backlog immediately through a time-bound camp with joint scrutiny teams, which simultaneously relieves the public and removes the leverage.
• Build evidence properly — refer to the Anti-Corruption Bureau or Vigilance for a trap and a discreet audit of approved files against dates of receipt. Chronological anomalies are documentary evidence already on record.
• Act on the ringleaders with proof, not on the office by suspicion; ban tout entry to the Collectorate and regulate access.
• Protect the honest majority — a confidential grievance channel, an explicit assurance that compliance will be backed, and public recognition for staff who cooperate. Their fear is the network’s real asset.
• Engage the union early on the process reform, where interests align, so that later individual action cannot be reframed as an attack on workers.

Conclusion

Seema’s task is not to win a confrontation but to remove the conditions that made the racket possible while holding its organisers to account. Transparency does most of the work: corruption of this kind lives in the gap between when a file arrives and when anyone can see it. Our note on measures to tackle corruption covers the systemic instruments.

The dilemma

Vikas must reconcile two duties that the system has placed in opposition: fiscal probity, which biometric authentication was introduced to protect, and the right to food of people whose fingerprints fail for reasons entirely outside their control — manual labour, ageing, childhood, and absent connectivity. The asymmetry matters: a leakage is a financial loss to the state, an exclusion is a hungry family.

Options open to Vikas

• A. Strict enforcement. Merits: no leakage, full rule compliance, no personal risk to him. Demerits: denies food to genuine beneficiaries, engages Article 21, and inverts the scheme’s purpose.
• B. Leave it to distributor discretion. Merits: nobody is turned away, workable today. Demerits: unauditable, invites ghost entries and quiet diversion, and makes the beneficiary dependent on the dealer’s goodwill — replacing a rights-based entitlement with a favour.
• C. Suspend biometric authentication in the district. Merits: immediate relief. Demerits: exceeds his authority, removes a genuine anti-diversion control, and is a blunt response to a targeted failure.
• D. Institutionalised exception handling with an audit trail — authentication remains the norm, and documented, verifiable alternatives operate where it fails.

The recommended option: D

The false choice in this case is between technology and discretion. The right answer is structured discretion that leaves a record.

• Multi-modal authentication — iris and face where fingerprints fail, which specifically addresses manual labourers and the elderly.
• OTP-based and offline authentication, with store-and-forward for low-connectivity days, so bandwidth is not a condition of eating.
• A mandatory exception register. Where a distributor uses personal identification, it must be recorded with the reason, photograph and signature, and reconciled later. This preserves both relief and auditability — the two things options B and A each sacrifice.
• Nominee and family authentication so an ailing beneficiary is not required to appear personally.
• Statutory backing already exists. Section 7 of the Aadhaar Act and the Supreme Court’s Aadhaar judgment (2018) both require that no genuine beneficiary be denied a benefit for authentication failure; the exception is not a favour Vikas grants but a duty he enforces.
• Monitor exclusion, not just leakage. Track authentication failure rates and undelivered entitlements by shop, and treat a high failure rate as a system defect rather than beneficiary fault.
• Escalate the design flaw upward with district data, since a fix at state level serves every similar district.

The ethics of technology as gatekeeper

• Error asymmetry is the central principle. A false positive costs the exchequer money; a false negative costs a family food. These are not commensurable, so the system should be tuned to tolerate the first rather than the second.
• Burden of proof is inverted. The beneficiary is made to prove her identity to a machine and bears the cost of the machine’s failure — a reversal of how entitlements should work.
• Technology is not neutral. Fingerprint failure correlates with manual labour, age and poverty, so an apparently uniform rule falls hardest on the intended beneficiaries.
• Dignity. Repeated failed attempts before onlookers, and repeat visits to the shop, impose a humiliation that no leakage statistic captures.
• Authentication is not verification of entitlement. Conflating “the machine did not recognise her” with “she is not entitled” is a category error that has cost lives.
• Accountability must follow the decision. When an algorithm or device denies a benefit, a human official must remain answerable and a grievance route must exist.

Conclusion

Fiscal probity and empathy are not opposed here, because the scheme’s purpose is the food, not the fingerprint. A control that prevents leakage by producing starvation has not achieved probity; it has redefined success. Vikas should keep the technology, refuse to let it be the last word, and make every exception visible enough to be audited. Our note on direct benefit transfers covers the wider design debate.

The ethical issues to be addressed

• Impartiality has already failed. The rivalry has entered the administration and police, so the state is no longer a neutral arbiter but a suspected participant. Restoring perceived neutrality is the precondition for everything else, not a later refinement.
• Duty to protect life first. Deaths are occurring and supplies are cut. The immediate obligation is to both communities equally, and visibly so.
• Equal treatment under conditions of unequal power. Formal even-handedness can entrench an existing imbalance in land, resources and representation; but visible tilting destroys legitimacy. Holding both is the hardest part.
• Justice against order. Quick restoration of order may require deferring accountability for the violence; deferring it too long tells victims their loss does not count.
• Use of force. Requisitioning CRPF is ethically defensible where local forces are compromised, but force must be proportionate, non-discriminatory and accountable.
• Truth and rumour. In communal conflict, misinformation kills. Managing information honestly is a life-saving duty, and suppression usually backfires.
• External exploitation. A neighbouring country may fuel insurgency, which raises the stakes but must not become a pretext for treating either community as suspect.
• Relief camp equity. Separate camps risk unequal provisioning becoming evidence of bias.

Options open to Rajan

• A. Rely on the local administration and police. Merits: fastest, uses local knowledge. Demerits: they are compromised; any action they take will be read as partisan by one side and will deepen distrust.
• B. Impose a heavy security clampdown first — curfew, mass preventive detention, force-led control. Merits: may stop violence quickly. Demerits: alienates both communities, risks excesses, and buys order at the cost of the legitimacy needed for a settlement.
• C. Move straight to political negotiation on land, resources and representation. Merits: addresses root causes. Demerits: impossible while violence continues and supplies are cut; talks under duress fail.
• D. Sequenced approach — neutral force to secure the humanitarian corridor, visible rebalancing of the administration, equal relief, then structured dialogue.

The recommended course: D

Rajan’s own instinct — open the highway first — is correct, and the sequence matters as much as the content.

• Requisition CRPF for the corridor and camps. A neutral, non-local force is the right instrument precisely because local units are compromised. Frame it publicly as neutrality, not escalation.
• Reopen the highway as a humanitarian act, serving both communities’ food and medical needs. This is the least contestable first move and demonstrates impartiality by deed rather than assurance.
• Rebalance the district team visibly. Move officers with demonstrated partisanship out of operational roles, and constitute mixed teams so no decision has a single-community face.
• Equalise relief measurably and publish per-camp provisioning data. Publication is what converts an assurance into a verifiable fact.
• Investigate violence even-handedly. Register cases from both communities on the same standard; selective FIRs are the fastest route to losing both sides.
• Build a joint civilian structure — a peace committee with respected figures from both communities, women’s representatives and traders who share an interest in reopening the economy.
• Counter rumour with a single verified information channel, issued in both languages, rather than a blanket communications shutdown that fuels speculation.
• Escalate the structural questions. Land claims, resource allocation and political representation are beyond a DM; document them and refer them upward while managing the immediate crisis.

Protecting his non-partisan image

Perceived neutrality is built from symmetry, transparency and distance: identical treatment of parallel incidents, published data on relief and arrests, use of a non-local force, mixed teams, refusal of hospitality or private meetings with leaders of either side, and a written record of every major decision with reasons. Our note on communal violence covers the administrative playbook.

Conclusion

Rajan cannot resolve conflicting land claims in a fortnight, and should not try. What he can do is make the state credible again — by acting first where both communities’ interests coincide, treating parallel wrongs identically, and putting his reasoning on record. Impartiality in a polarised district is not a stance one announces; it is a pattern that accumulates from decisions that can be checked.

The decision

Production constraints permit one order. The question is framed as economics against ethics, but on these facts the two converge — and recognising that is the strongest move available.

The two buyers, compared

• Country A — sound technology base and a planned R&D collaboration for next-generation weapons; no security alliance, so no third-party leakage route; a defensive requirement against a troublesome neighbour; large order on long-term loan.
• Country B — military spending already displacing human development allocations; hosts a superpower’s base and receives its grants; NPT non-signatory holding tactical nuclear weapons and delivery systems; supports guerrilla forces abroad; smaller order with some advance payment; simultaneously negotiating with another supplier.

The ethical issues

• End-use and diversion risk. B’s support for guerrilla forces abroad is the single most serious fact in the case: weapons supplied to a state that arms non-state actors may reach those actors. This is a direct end-use violation risk, not a speculative one.
• Complicity. A supplier who foresees misuse and proceeds shares moral responsibility for the consequences.
• Development displacement. Selling to a state that already diverts human-development budgets to arms deepens harm to its own population.
• Proliferation and technology security. B is outside the NPT and hosts a foreign military base; sensitive technology could be exposed to both.
• Regional stability. Long-range artillery and missiles into an already militarised setting can trigger an arms spiral.
• The “someone else will sell anyway” argument. This is the classic rationalisation and it is ethically empty: the availability of another wrongdoer does not license the wrong. It also happens to be commercially weak, since B is negotiating elsewhere regardless.

Options before Ajit

• A. Sell to Country B. Merits: advance payment, quicker cash, possible goodwill with the economic bloc during FTA talks. Demerits: end-use and proliferation risk, reputational damage, technology exposure, and complicity in regional destabilisation.
• B. Sell to Country A. Merits: R&D partnership with long-term capability gain, no alliance-leakage route, defensive end-use, larger order building scale. Demerits: long-term loan defers revenue and carries credit risk.
• C. Refuse both pending capacity expansion. Merits: avoids all risk. Demerits: forfeits revenue, employment and strategic relationships for no ethical gain, since A raises no serious concern.
• D. Split the order. Merits: appears balanced. Demerits: production constraints forbid it, and it would take on B’s risks anyway.

The recommendation: Country A

Option B should be selected, and the justification should rest on both grounds rather than on sentiment alone.

• Ethically — a defensive requirement, no record of arming non-state actors, no proliferation concern, and no foreign base through which technology could leak.
• Strategically — the R&D collaboration converts a one-time sale into a long-term capability partnership, which is worth more than one advance payment.
• Commercially — a larger order supports production scale; the long-term loan can be de-risked through export credit guarantees and sovereign instruments rather than treated as a reason to refuse.
• Reputationally — a supplier known for disciplined end-use scrutiny becomes more attractive to serious buyers, not less.

Balancing economic and strategic interest with ethics

• Institutionalise, do not improvise. Written end-use certificates, no-retransfer undertakings, and post-shipment verification rights turn a personal judgement into policy.
• Apply the existing framework — SCOMET controls, the Arms Trade Treaty criteria on diversion and human rights, and UN embargo checks — as a mandatory gate rather than a formality.
• Country risk assessment covering conflict involvement, non-state actor support, human rights record and treaty status, with inter-ministerial sign-off from MEA and defence.
• Sequence the questions. Legality first, then end-use risk, then strategic value, then commercial terms. Reversing that order is how bad sales get approved.
• Record dissent. If overruled at the political level, Ajit’s duty is a written, reasoned note — the decision is the government’s, the record is his.

Conclusion

Ajit should recommend Country A. The case is instructive precisely because the ethical choice is also the better strategic one: a partnership that builds capability beats a payment that buys risk. Where the two genuinely diverge, due diligence at every stage — which Ajit already knows to be pivotal — is what keeps a defence-export programme from becoming an instrument of somebody else’s war. Our note on weapons and international regulation covers the treaty landscape.