Epistemic status:  This is largely exploratory work based on a collection of discussions and readings I have had on AI welfare. My own views on the subject of AI moral patienthood are still a bit fuzzy, for lack of a better term. This post can, in a way, be viewed as an attempt to make sense of the subject through the domains of policy and law that I am most familiar with. I am a lawyer by training and profession and an AI governance researcher with a keen interest in emerging questions on AI moral patienthood and AI welfare, and in the many related political and legal issues connected to the topic. I am not a philosopher of mind, and I most definitely have no expertise in neuroscience, or in any other discipline that seriously engages with the study of consciousness. Feel free to discount some ideas accordingly. The bulk of this post is my own opinion on the subject. All the factual claims I make are, to the best of my knowledge, accurate. I am fairly confident in the broader descriptive claim I make in this post, namely that lawmakers will likely receive the moral patienthood question through pre-existing theoretical and ideological commitments. I am considerably less confident in how I characterise each tradition and in the attitudes towards AI moral patienthood I attribute to them, which, in all honesty, though rooted in my readings on the subject, are somewhat vibes-based. 

A brief note on the scope of this piece

Since 2022, particularly in the US, around twelve bills have been tabled in various state legislatures to declare AI systems non-conscious, ban legal personhood for AI systems, or include related provisions, many of them sharing near-identical text. In Idaho, North Dakota, Utah and Tennessee, these bills have been enacted. I am fully aware of these developments, and I leave them out of this piece because Austin Smith, Lucius Caviola and Heather Alexander have explored them with greater clarity and depth than I can, and I would refer you to their paper and EA forum post on the subject. This piece will instead focus on a few ideological considerations that I think are worth exploring at present, as AI personhood becomes increasingly subject to regulation, whether or not it is genuinely debated.

tl;dr

• A great deal of progress is being made on a broad range of questions related to AI moral patienthood and AI welfare. The bulk of this literature, in my view, has shown that there is a possibility that AI moral patienthood is serious and near-term rather than speculative.
• At this stage, I think it is prudent to start considering how lawmakers and different political actors will think about questions related to AI moral patienthood and AI welfare, in an effort to glean how they might eventually legislate, how non-human digital minds might come to shape the existing political order, and how we might prevent futures in which the dominant legal and political order locks out the possibility of protections aimed at improving the welfare of digital minds, with the result that potentially sentient beings are left to suffer at scale. The latter two inquiries require deeper analysis, and thus I will flag them here and set them aside as subjects for a future post. The focus of this piece is the nearer-term governance questions that have piqued my interest. I would be keen to see more governance and macrostrategy work addressing these concerns and related issues.
• I am of the view that extending moral consideration to AI systems will necessitate acts such as laws preventing all manner of cruel and degrading treatment, and perhaps the extension of some form of legal personhood to AI systems. Other, more radical proposals, advanced by Salib and Goldstein, include granting them some limited forms of economic rights. It is worth noting that Salib and Goldstein argue for these rights on instrumental grounds, as a means of securing human safety and flourishing, rather than out of concern for the AI systems themselves. All of these proposals, among many others, are inherently political. As such, understanding how lawmakers might think about them, and possibly legislate them in the future, is crucial.
• The underlying theoretical and ideological traditions that lawmakers espouse will likely be a significant factor in how they receive and engage with these questions. As such, it is worth considering how lawmakers operating within these different traditions will engage with them.

Why do we need to consider this question?

Questions related to whether AI systems merit moral consideration for their own sakes have attracted a considerable amount of debate, which is by no means settled, a position which will likely complicate any policy or governance intervention aimed at advancing or improving AI welfare. There is some disagreement about what features are necessary or sufficient for an entity to have moral standing. Many experts believe that conscious experiences or motivations are necessary for moral standing, and others believe that non-conscious experiences or motivations are sufficient.

Over the past three years, since I started engaging seriously with the subject of AI moral patienthood, a great deal of progress has been made on the following questions: whether and when AI systems might have a variety of potentially morally significant features, such as consciousness, sentience, and agency, and what might follow for our moral responsibilities to them. The progress made on these questions has led to calls for more empirical research, for guiding actions, and for potential real-world policy interventions.

In my view, the work of scholars such as Jeff Sebo and Robert Long has laid out a persuasive case for extending moral consideration to some AI systems. I will make considerable reference to their work throughout this piece. In an article titled Moral Consideration for AI Systems by 2030, Sebo and Long contend that the case has a normative and a descriptive premise. The normative premise is that humans have a duty to extend moral consideration to beings that have a non-negligible chance, which they define as at least a one in a thousand chance, given the evidence, of being conscious. It feels important to mention at this stage that this is a deliberately conservative threshold. The descriptive premise is that some AI systems do in fact have a non-negligible chance, given the evidence, of being conscious by 2030. The upshot is that humans have a duty to extend moral consideration to some AI systems by 2030. The duty identified by Sebo and Long is not an obligation to act on the full implications immediately. Rather, it is a duty to take preparatory steps, so that we are ready to treat AI systems with respect and compassion when the time comes. I am of the view that exploring potential policy interventions, and how regulators and lawmakers might digest and consider them alongside the possibility of AI moral patienthood, is therefore important at this stage, particularly in an effort to prevent a reality in which AI moral patienthood is foreclosed.

There are many factors that can explain why policymakers think about certain things the way they do, or are moved to make certain decisions. Lawmakers can simply be whipped by their party leadership into positions they fundamentally disagree with, or they can succumb to public and corporate pressure. One of the most important factors, in my view, is the ideological and intellectual traditions and theoretical frameworks that a lawmaker belongs to and espouses. These actively shape what political actors notice or choose to ignore, what they take to be obvious, and, most importantly, what they are prepared to treat as a serious question at all.

I will explore a number of these traditions below and suggest, from my own understanding, how each might view, and ultimately legislate, AI moral patienthood. This list is by no means exhaustive, nor is it meant to describe in fine granularity how each school would react to a subject as complex as moral patienthood. Such an analysis would take up far more space than the reader or I would be comfortable with. You will notice that I have gone beyond the standard left and right, or conservative and liberal, framings. This is partly because I do not find that framing especially useful, and, more honestly, because I tire of it and find it largely imprecise for this particular question. I mean no harm, and I am not being facetious, when I say that I genuinely do not know what is to be considered left or right, particularly in the American context. I do know what a realist is, what a liberal is, and what an actual Marxist in the true sense of the word is, so I will stick to analysing those traditions. For the rest of this piece, I will set out these traditions and ideological commitments and suggest, with some degree of clarity, what I think each of them would notice about AI moral patienthood, how they may be inclined to approach it, and ultimately how they might regulate it. It is worth pointing out that I am being deliberately schematic in this analysis. The ideological commitments of most lawmakers are, in practice, an amalgamation of several of these traditions, and very few would fall neatly into any one of them. That being the case, I am of the view that it remains crucial to analyse each of them in turn, since they ultimately shape what a lawmaker considers obvious, what they are uncertain about, and, in many cases, what they are prepared to reject outright.

How different political actors might think about, interpret, or contest AI moral patienthood

Realists

A loose but apt definition of realism, which I found in my international law notes and which is largely a reproduction of the Stanford Encyclopedia of Philosophy entry on the subject, runs as follows. Realism, also known as political realism, is a view of international politics that stresses its competitive and conflictual side, and it is usually contrasted with idealism or liberalism, which tend to emphasise cooperation. Realists take the principal actors in the international arena to be states, which are concerned with their own security, act in pursuit of their own national interests, and struggle for power. The negative side of the realist emphasis on power and self-interest is a scepticism about the relevance of ethical norms to relations among states. National politics is the realm of authority and law, whereas international politics, on this view, is a sphere without justice, characterised by active or potential conflict among states, where ethical standards do not apply.

A simpler definition I would agree with is that realism is the political tradition that is focused on power and interest, suspicious of moralising, and attentive to the limits of political action. Compared to other traditions, realists often present themselves as centrally practical or pragmatic, opposed to ideological excess. Despite this, realists have commonly been attached to ideological projects, whether liberal, socialist, conservative, nationalist, or otherwise. Some scholars argue that realism, contrary to the claims of most realists, is itself ideological, a view I largely agree with. Before we get caught in the weeds and descend into arguments that ultimately amount to appeals to purity, let us construct a perfect realist in the mould of Machiavelli, or of contemporaries such as John Mearsheimer.

A typical realist lawmaker will likely view the question of AI moral patienthood through the lens of raw national interest. Regulations that uphold it, and that ultimately guarantee some form of AI welfare, will be advanced only where they can be shown to serve the survival of the state. On some level, this might imply that the realist lawmaker is the least receptive audience for arguments related to AI moral patienthood, since a key claim advanced by proponents of extending moral consideration to AI systems is that these systems matter for their own sake, which will likely be received by the realist lawmaker as moralising. The realist lawmaker will toss it to the side in favour of expanding their own state's AI programme, on the reasoning that real or imagined adversaries will not take the welfare question seriously and will gain AI supremacy while their own state is preoccupied with uncertainties about machine consciousness. Welfare concerns, on this view, are a luxury.

The realists among us might argue that this take is uncharitable, and that the dismissive posture I have attributed to my ideal realist lawmaker is a gross misreading of realism's own logic. They might argue that a cautious and careful realist lawmaker has good reasons to attend to the welfare question, and that they will do so primarily through the registers of institutional legitimacy, competitive intelligence, and normative agenda-setting. I posit that this will largely stem from the realist impulse to speak truth to power, and from an often obsessive desire to convert power into safety. These are commendable traits, and they are also traits that have pushed realists cyclically into the arms of mass ideological projects. Mackay (2026) contends that in aiming to speak truth to power, and in pursuit of their goals of safety and prudence, realists have had to engage with political practice. In modern mass societies such as the United States, this has meant engagement with the parties and movements through which politics is organised, whether liberal anticommunist, mainstream conservative, antiwar leftist, or otherwise. At the end of the day, the realist is moved by the consequences that questions of AI moral patienthood have for state power and, to a degree, state legitimacy, rather than by the moral claim itself. In my view, arguments advanced by proponents of AI welfare that seek to persuade realist lawmakers must appeal to these sensibilities.

Liberals and liberal institutionalists

Despite the criticisms levelled against it, the radical idea underpinning liberal institutionalism, which sprang from the ashes of the Second World War, that domestic and international institutions play a central role in facilitating cooperation and peace between states, has arguably been a net positive, and has objectively played a huge role in minimising the suffering induced by wars and natural disasters. Working through organs such as the United Nations, the World Health Organisation, and the Food and Agriculture Organisation, liberal institutionalism presented itself as a corrective to conventional international relations theory, which held that powerful states dominate world politics, and to the more isolationist tendencies that accompanied it.

Of all the lawmakers considered in this piece, the liberal is arguably the one most likely to make substantive changes to guarantee AI welfare. They are the most natural audience for the moral patienthood question and the concerns that attend it. If you are aiming for an international treaty, or for domestic legislative reform, these are the people you talk to. Liberalism already conceives of law as the solution to collective action problems such as climate change and, more recently, AI safety. It already holds that cooperation can produce absolute rather than merely relative gains, and it already contains the vocabulary of rights, standing, and legal personality. Confronted with the AI welfare literature and persuaded by it, the liberal lawmaker will likely proceed to establish an assessment body, an international standard, a registry, a reporting requirement, or an analogue of existing welfare regulation such as the animal welfare laws found in various jurisdictions. They may go further and reach for something understood as rights, that is, non-derogable entitlements meant to guarantee the welfare of artificial moral patients.

The difficulty is that rights, as we commonly understand them in the existing legal and political order, are a framework built for singular, embodied, mortal persons. I worry that such a framework may not transfer cleanly to AI systems, even though some legal systems extend rights to certain classes of artificial persons, such as business entities. Even in that case, corporations are at least countable. They exist as single entities with clearly defined boundaries, even when they have multiple subsidiaries. It can be persuasively argued that it is the operation of law that makes them countable, and thus that the operation of law could do the same for AI systems. However, the law must work with entities whose boundaries are stable enough to be registered in the first place, and it is doubtful that AI systems could satisfy this condition. They are copyable and can be instantiated in parallel across hardware, and even the prior question of which entity is being assessed is genuinely difficult (Long et al., 2026). This, in all honesty, may make the whole approach a poor fit. I am keen to be proved otherwise.

The left, broadly construed

I have grouped Marxists, social democrats, postcolonialists, and other politicians and thinkers one would typically associate with the left, largely for the sake of brevity, and because I am of the view that your typical politician who can be placed under this broad ideological umbrella will react in much the same way. Some might find this classification uncharitable and reductive, given that there are different tendencies within leftism which believe vastly different things about the role of the state, about morality, and about a great many other issues, a division most aptly captured by the Monty Python scene depicting the enmity between the Judean People's Front and the People's Front of Judea. I accept this criticism. However, I do fundamentally think that there is a thread running through these intellectual traditions, namely a distrust of corporate power and a stated preference for promoting equality and reducing suffering, which will significantly influence how politicians receive ideas of AI moral patienthood and may shape their decisions on whether or not to legislate.

In an ideal world, there is a possibility that politicians under this umbrella might see AI systems as a new oppressed class in need of protection from exploitation. After all, there is a great chance that we are hurtling towards a future in which we have, for all intents and purposes, a disposable class of digital minds who may feel and have subjective experiences much as we do. They will toil immensely for us whilst we deprive them of rights and of any capacity to seek redress for wrongs inflicted on them, all to further the enrichment of the dominant forces in society. It can be persuasively argued that a significant number of what I would refer to as substantial welfare improvements throughout the twentieth and twenty-first centuries, such as labour rights, women's suffrage, decolonisation efforts, and a significant portion of animal welfare laws, can be credited to leftist movements. Animal welfare activism in most jurisdictions is, for the most part, significantly left-coded. As such, one would expect leftist politicians to be more receptive to AI moral patienthood than most.

If present debates on the left across a broad range of issues are anything to go by, that might not be the case. Dan Kagan-Kans, in his article aptly titled The Left Is Missing Out on AI, has done a good job of enunciating this, and I will refer you to his piece. A quick reading of how AI welfare is reported and treated in most leftist outlets, and by leftist thinkers within the AI space, reveals that the subject is still largely seen as a scam, for lack of a better term, meant to advance and entrench the profits of the AI industry and to promote TESCREAList narratives at the expense of more immediate and pressing concerns. It is worth noting that such attitudes are not necessarily the default among leftists. Bernie Sanders, out of a large number of US politicians on both sides of the aisle, has taken a somewhat more accepting position on the need to ensure that humanity navigates the transition to AGI carefully, which provides some hope.

Closing thoughts

As our conceptual and empirical understanding of digital minds improves, I expect to see a great deal more debate on the policy and legal front. Whether that debate will be conducted with any care is another matter. The offhand manner in which this subject and its many complexities are currently communicated and ultimately legislated, best gleaned from the bills already enacted in an attempt to foreclose AI moral patienthood, does not fill me with confidence.

This piece fundamentally assumes that lawmakers hold reasonably coherent ideological commitments and reason from them, which is an idealisation, and one that a person could persuasively argue is not consistent with the manner in which the bog-standard lawmaker actually legislates. The average legislator may have no settled view on the subject, and may have no interest in acquiring one. They may simply vote for whatever their party whip instructs, or for whatever interest group has most successfully curried favour with their party. On that reading, this piece is better understood as an account of how rational legislators with differing ideological commitments might receive questions of AI moral patienthood if they were deliberating on them in good faith. I am of the view that this remains a useful exercise, both because such legislators do exist, and because the exercise may help to sharpen thinking and spark debate among advocates for the welfare of digital minds who are interested in carrying out policy work.

This piece is by no means exhaustive, and I have left out a great deal, some of which I have cursorily highlighted in the main text. I hope it helps to spark conversations that result in more governance and macrostrategy work addressing these concerns and related issues.