Governing to prevent the worst suffering

From morphine denied to dying patients to autonomous weapons with no off switch, the case for putting suffering prevention at the heart of how we govern

Jonathan Leighton speaking at EAGxAustralia 2023

By Jonathan Leighton, Executive Director, Organisation for the Prevention of Intense Suffering (OPIS)

The questions policymakers tend to avoid are often the most important ones. How do you build governance systems around preventing suffering when bureaucratic inertia and short-term thinking dominate the room? How do you talk about compassion in a security briefing without getting laughed out of the building? And what does any of this have to do with AI?

These are questions I’ve spent years working on, through OPIS’s advocacy on pain relief access, through my books, and most recently through our Compassionate Governance guide. What follows is my attempt to address them directly, for an audience that I think is ready to take them seriously.

Compassion is not naivety

Let me be clear about what compassion actually means as a governance principle. It is a motivational stance, an openness to the suffering of others, with the goal of alleviating it. It does not imply blind trust or weakness, or require making unilateral moves that won’t be reciprocated. There is nothing inherently naive about compassion as a driver of governance.

That said, a security minister might not be the first person to try to persuade, as their focus is more narrowly on deterrence and the ability to respond to military threats. But prioritizing compassion has concrete dimensions that even security-conscious leaders could embrace. It can become an essential aspect of governance culture, starting with domestic policies that address suffering within a jurisdiction and build trust between citizens and their governments. It can signal to adversaries a willingness to address issues collaboratively and transparently. It provides an effective and proven angle for conflict resolution, one that addresses underlying needs rather than merely assessing the strength of adversaries’ stated negotiating positions.

On a more ambitious level, compassion can provide a common purpose capable of uniting populations across borders, and it can serve as the foundation for frameworks of global cooperation that use well-designed incentives to achieve compassionate aims. There are good reasons to believe that compassionate governance can actually lead to much greater security.

The gap in AI governance

Current attempts at AI regulation are clearly insufficient for preventing powerful superintelligence from emerging, with potentially catastrophic consequences for human well-being. Stable technical alignment is difficult, at the very least, which means maintaining responsible human control over AI systems is essential. But as we highlight in a chapter of the Compassionate Governance guide, even successfully aligning AI with our society’s current values would still be catastrophic, given the amount of preventable suffering in humans and animals that is currently tolerated.

In my book The Tango of Ethics, I wrote:

“Surely we would want to encode a powerful AGI with the most utopian ethical framework we can, rather than lock in the prevailing ethics of the early twenty-first century.” We need explicit concern for suffering to be embedded in AI systems, through constitutions, training, and the use of suffering metrics and forecasting to assess AI’s possible impact on suffering.

What extreme pain teaches us about governance

The fact that governments don’t explicitly prioritize suffering prevention has a direct consequence that some of the worst forms of suffering are neglected. In lower- and middle-income countries, most people suffering from severe pain, including from terminal cancer, are unable to obtain morphine or similar opioids, despite morphine being one of the few medications that can provide adequate relief, and despite it being on the WHO’s Essential Medicines list. Cost is not the main obstacle. The barriers are unjustified fears about dependency or diversion to black markets. The experience of sick people crying out in pain as they lie on a bed in a dark room in a rural village seems not to be tangible or visible enough for most decision-makers to act.

With cluster headaches, the situation is a little different. It is an excruciating condition for which certain psychedelics are often more effective than any other treatment, but also illegal in most jurisdictions. Large-scale clinical trials are difficult and expensive to run, and there is little incentive for pharmaceutical companies to run them for unpatented chemicals. One of the main problems is that most countries lack adequate compassionate use provisions that would allow doctors to prescribe these substances to patients in need even without formal pharmaceutical approval. Drug policy itself is often uncompassionate and counterproductive, criminalizing users instead of treating drug use as a public health issue, and threatening patients who self-medicate with prosecution.

The current governance system in most countries isn’t structured to treat extreme suffering like the emergency that it is. This needs to change.

Non-human suffering in governance

Non-human suffering in conflict is a rarely discussed issue, and I’m glad to see it taken seriously by the editors of this issue, if only to raise awareness of one more dimension of how human activities harm other beings. There are, of course, much larger-scale, systematic ways that humans cause suffering to animals, particularly in the food industry. This includes not just factory farming, but the hundreds of billions of fish and other marine animals raised and harvested for human consumption, who suffer agonizing deaths from suffocation. It is important to recognize that the suffering of fish is not intrinsically less important than that of marine mammals we feel greater affinity for.

Suffering caused by military activity might therefore seem like a lower priority for intervention. But achieving impact in that domain might in some ways be easier, requiring not a shift in consumer habits but a high-level decision within a military to incorporate animal welfare into operational guidelines.

Realistically, we are still very far from that goal, given how little attention is paid to non-human animal suffering in international law and national legislation, and how difficult it is to get militaries to avoid causing human suffering once war breaks out. My greater hope is that we move toward a world where violent conflict is no longer perceived as acceptable or necessary. In that world, we can also build better, voluntarily agreed international enforcement mechanisms that protect both humans and non-humans from harm.

Trust, AI targeting, and conflict

Armed conflict already means trust between parties has broken down. Adherence to prior commitments, such as non-use of certain weapons or technologies, relies mainly on the incentives in place and the consequences for a country’s international relations if those commitments are broken. These considerations apply when nuclear powers engage in conventional warfare, or when states refrain from using banned weapons such as cluster munitions or white phosphorus.

The use of AI targeting systems is far more opaque and harder to monitor. It seems unlikely, at the moment, that a treaty will be able to limit their use for target identification, which in principle can both reduce and increase risks depending on how cautiously such systems are deployed. The main incentive for a country to be more careful is the international fallout that follows mass civilian casualties.

My hope remains that we can more effectively prevent armed conflicts from happening altogether, through more effective global governance grounded in a genuine addressing of needs, where trust is built steadily before conflicts have a chance to break out. I don’t think we can afford to be any less ambitious than that.

---

Who Governs AI When the World Is Divided

---

Power, AI, and who gets left out

Global AI governance is extremely challenging. Corporations and governments stand to benefit in the short term from advances in AI systems at their disposal, and there is a kind of arms race, fed by fear of adversaries developing superintelligence first, even though everyone stands to lose in the long run. But the alternative to effective regulation is not acceptable.

The decision to produce our Compassionate Governance guide was motivated in part by the realization that, if we fail to achieve closer global cooperation, it may soon be too late to address the enormous risks of uncontrolled AI, or the entrenchment of totalitarianism in countries that use powerful AI to suppress dissent. We desperately need more ambitious, more comprehensive approaches to these issues. Chipping away at the problem in a piecemeal fashion is unlikely to be sufficient. And there is currently a severe absence of trusted ethical leadership in the world that understands the need to seek close cooperation, at a time when that leadership is most urgently needed.

Democratic reform before it’s too late

I have a strong belief in the expanded use of citizens’ assemblies at all levels, including the creation of an effective global citizens’ assembly.

The key advantages are that issues get deliberated on calmly and transparently, with expert input, by ordinary people who don’t carry the conflicting interests of career politicians. This allows for greater consideration of ethical concerns, longer-term societal interests and, at the global level, the perspectives of citizens from other countries. It is potentially a much better system for achieving global cooperation, and I hope to see citizens’ assemblies become an integral and powerful component of governance structures. Giving them greater power should make democratic systems more resilient to disruption and manipulation.

Other useful democratic reforms would include shifting away from first-past-the-post systems toward ranked choice or approval voting, ensuring that smaller parties with genuinely good ideas can attract votes without those votes being wasted. This could make it easier for progressive ideas to be implemented, including stronger measures to restrict the misuse of AI.

Whether such structures and changes can be implemented quickly enough to avert AI risks is uncertain, to say the least. AI is advancing at a breathtaking pace, while democratic reforms are slow to be agreed on and implemented. Politicians would need to commit to instituting reforms that could see them lose some of their own power. In the United States right now, the ruling party is moving in the opposite direction, stripping some voters of their democratic voices. The timing makes this approach to politics especially damaging, given current AI timelines.

The speciesism problem in policy

Ultimately, we would want the prevention of non-human suffering to be embedded in constitutions and formal decision-making mechanisms, just as we want for humans. Because citizens’ assemblies are being used more widely and hold real promise as a democratic innovation, an important step forward would be to ensure that animal well-being becomes one of the guiding principles for their deliberations, giving animals a so-called voice at the table. This is something we are actively advocating for.

In the meantime, progress happens on an issue-by-issue basis, sometimes at the local level. Amsterdam and other Dutch cities recently banned meat advertising, even if the primary motivation was environmental rather than animal welfare. Given the massive numbers of chickens and other small animals kept under horrendous conditions on factory farms, one would want to see governments taking action to ensure much higher welfare standards, even if it means higher consumer prices. But such policies might only gain public acceptance if people don’t feel they are being asked to make major sacrifices. It may be that only competitively priced cultured meat can create the conditions for governments to legislate a dramatic reduction and eventual elimination of the cruelty of animal agriculture. Political will ideally needs the support of technological innovation and a shift in public sentiment.

The accountability gap in public-private AI deals

As a general principle, when decision-making is not guided by compassionate values, there is always a greater risk of suffering as a result. Governments’ main responsibility is to protect the well-being of the humans and animals living within their jurisdictions, with concern as well for those beyond their borders. Contracts with the private sector are not necessarily a problem, but the question is always: to what end? Are possible efficiency gains being used to increase well-being, or mainly to increase shareholder value?

New public-private partnerships should always be evaluated by politically neutral bodies, based on careful benefit-risk analysis that ideally includes an assessment of predicted impact on suffering, using appropriate metrics.

More broadly, what is needed is a formal entrenchment of values, including compassion, reason, and transparency, and the development of a governance culture aligned with those values. Governance reforms that shift power away from career politicians toward independent deliberative bodies that weigh evidence objectively would provide a real check against corrupted decision-making and undue influence from tech companies.

Conflict resolution when machines fight

A world where armed conflict persists is one where compassionate governance has not yet been sufficiently implemented at scale, and we are left in damage control mode. But if humans no longer have the power to implement conflict resolution mechanisms because AI systems are not functionally accountable to them, humanity has lost a battle.

Establishing a verifiable and enforceable ban on fully autonomous weapons is essential. More broadly, we must ensure that powerful AI systems retain humans in the loop who can intervene and shut them down if necessary. While humans are disturbingly prone to conflict, we are still far from the point where we can voluntarily relinquish our future to AI systems more ethical and dependable than we are.

Unless and until AI systems are stably aligned with deeply compassionate values, we must do all we can to retain meaningful human control over them, and over our future.

---

Jonathan Leighton, PhD, is an ethics strategist and the Executive Director of the Organisation for the Prevention of Intense Suffering (OPIS), a Swiss think-and-do tank he founded to promote compassionate ethics and policymaking. He is the author of The Tango of Ethics: Intuition, Rationality and the Prevention of Suffering and The Battle for Compassion: Ethics in an Apathetic Universe, as well as OPIS’s 2025 publication Compassionate Governance: A Strategic Guide to Preventing and Alleviating Global Suffering. Learn more about the author: https://www.jonathanleighton.org/

The views expressed in this article are those of the author and do not necessarily reflect the views of Techplomacy Magazine or the Techplomacy Foundation. Articles may be republished in full, without alteration, with credit to Techplomacy Magazine (magazine.techplomacyfoundation.org).

Series: Techplomacy Conversations™

---