What Success Looks Like For Animals

Crosspost from my Substack https://jlmasterson.substack.com/p/what-success-looks-like-for-animals 

Introduction

Humans kill billions of animals every single day.  The scale of suffering is so enormous as to be virtually incomprehensible. Animals are bred, mutilated, tortured, hunted, trapped, experimented on, beaten, starved, and kept in tiny cages. Despite this, it is possible that things could get worse for animals. In ‘What failure looks like for animals,’ Alistair Stewart and Niki Dupuis explore the ways in which a future with powerful AI could negatively impact animal welfare. This includes ideas such as AI wiping out the biosphere and anti-animal value lock-in. It is important and valuable to explore those prospects in order that we might be able to think more clearly about how to avoid them. Here, I will explore the other side of the story, examining the possible outcomes for animals if we get everything – or at least most things – right.

One thing we can be certain of is that AI is changing the world. According to Stanford's 2025 AI Index Report, the capabilities of AI are advancing rapidly, with benchmark performance improving by 18-67 percentage points in just one year, and the cost of using frontier AI models falling 280-fold in 18 months. Two Nobel Prizes were awarded for AI-related breakthroughs in 2024. AI systems now outperform doctors when it comes to complex diagnostics, power over 150,000 self-driving vehicle rides per week, and are deployed by 78% of organisations worldwide. With $109 billion in private investment in the US alone, and governments committing to invest hundreds of billions more in the future, this rapid growth shows no sign of slowing down. These changes are already impacting non-human animals. Precision livestock farming involves the use of sensors, surveillance, health monitoring tools, cameras, behavioural pattern recognition and more to revolutionise livestock farming. The global precision livestock farming market was estimated to be worth $7.5 billion in 2024 and is projected to reach $19.89 billion by 2033. Recording technology and collection of massive amounts of data are enabling advancements in interspecies communication. AI is being explored as a tool to accelerate development in the alternative protein sector, allowing us to optimise the extrusion process, improve crop production reliability, and more. By combining the trajectory of AI development with our already-evident willingness to use AI tools in human-animal interactions, it seems plausible that AI will radically change animal lives over the coming years.

As animal advocates, we have good reason to be apprehensive about what the outcomes of this shift will be. Thus far, humans don't have a great track record of using our power to make conditions better for animals – although a minority of people have, thankfully, done this. Nonetheless, the tone of this piece is deliberately optimistic. I have written before about the value of considering potential bright futures. Here, I will consider many bright possible futures, in the hope that this article can serve both as a reference and as a motivational tool in a field that can, due to the relentless present suffering, often feel dishearteningly bleak. The success for animals I refer to in the title does not require perfection. It requires that powerful AI systems are aligned in ways that structurally reduce suffering and extend moral consideration to animals. It is not clear to me that AGI is required for most of the applications explored here – advanced targeted AI could be sufficient in the majority of cases. The outcomes discussed here are by no means certain; if anything, some are highly unlikely unless we push exceptionally hard to make them a reality. Perhaps this piece will add some momentum to our push. That’s the hope.

Foundational Shifts

In later sections, I will look specifically at AI technologies that could change the lives of various groups of animals – farmed, wild, companion, and other. To start with, I will discuss broader and less species-specific shifts, such as AI-enabled moral circle expansion and widespread economic changes. These broad shifts shape the moral and economic terrain on which more specific interventions become possible. If, by any of the means explored below, the way that we think about animals fundamentally changes in a positive way, then the impact will be wide-reaching and multi-layered. 

Moral Circle Expansion

In The Expanding Circle, Peter Singer writes:

‘The circle of altruism has broadened from the family and tribe to the nation and race, and we are beginning to recognize that our obligations extend to all human beings… it is as arbitrary to restrict the principle of equal consideration of interests to our own species as it would be to restrict it to our own race. The only justifiable stopping place for the expansion of altruism is the point at which all whose welfare can be affected by our actions are included within the circle of altruism.’

Moral circle expansion is an attempt to extend moral consideration to a wider group, to expand the boundaries of who is considered worthy of such consideration. History has shown us that this is possible. The Enlightenment thinkers moved away from superstition and tradition, and towards reason and scepticism, with this shift in thinking ultimately laying groundwork for the abolition of slavery and torture, the end of the criminalisation of homosexuality in many countries around the world, equal rights for women, and an increasing concern regarding economic and social inequality. Already, far more people are concerned about animal welfare than were a few decades ago. One member of The Vegan Society explains that in 1969 ‘veganism was a niche lifestyle, often associated with fringe countercultural movements’. In 2025, approximately 25.8 million people participated in Veganuary in 2025, and 6.3% of UK adults are planning to follow a vegan diet in 2026[1]. Although moral circle expansion to include non-human animals does seem to be underway, it is possible that it will happen in an uneven, species-specific way. As Jacy Reese Anthis articulates, there has already been widespread condemnation of the killing and eating of dogs, but very little condemnation of the killing and eating of fish who are, in many relevant ways, similar. Whether it happens evenly or not, what would it take to accelerate the animal-focused moral circle expansion that already seems to be happening, and how could AI enable this acceleration? There are a few ways to consider.

Firstly, AI could allow for the optimisation of animal advocacy outreach. Animal organisations could utilise AI for targeted efforts, using data to tailor animal welfare messaging based on an individual’s preferences and values; campaigns could target specific people for whom the messaging would be most impactful by using available data. Personalised advocacy interventions offer highly relevant communications with messages and calls to action that could, theoretically, update in real time for individuals based on their browsing history and online interactions. This kind of intervention could also enable organisations to identify the most persuadable populations and focus their efforts there to maximise impact. By increasing the impact of advocacy campaigns and optimising opportunities to request support – whether financial, support for campaigns via signatures, pledges etc – this kind of advocacy could nudge our moral development forward regarding animals.

We could gain insights through AI-enhanced media, interactions with animal-friendly LLMs, and increased accessibility to expertise and information on animal welfare and cognitive capacity. AI could generate media that shows the perspective of animals, potentially increasing compassion in the viewer or reader. Synthetic data written from animal perspectives could be incorporated into LLM training, increasing the likelihood that these systems treat animals as moral patients. Adding in animal perspectives could make a substantial difference to the expected behaviours of LLMs, ‘causing them to actually empathize with animals which is valuable for alignment’. AI companies could incorporate these perspectives consistently in their training data. Compassion training has also shown promise as an intervention for moral circle expansion. A randomised controlled trial from 2024 found that by training people in Compassion Focused Therapy (CFT), they managed to cultivate the compassion motivation;  at the three month follow-up, researchers found that participants reported ‘significantly increased overall moral expansiveness, as well as increased moral concern towards all entity sub-groups’ (this included animals). CFT could be made widely available through AI, with LLMs optimising for this capacity and potentially even leaning into compassion-led interactions. Given the early results of the CFT trial, I am cautiously optimistic about the impact this could have. LLMs that also have animal-friendly values would force moral inconsistencies to the surface. Interactions could allow the AI to function as a kind of moral mirror, showing the contradictions inherent in treating cruelty to dogs and cats, for example, as abhorrent, but cruelty to pigs and cows as morally justifiable for the purpose of taste pleasure. Imagine that a user asks for a recommendation for a steakhouse. An AI could comply while also drawing attention to the fact that, earlier, they mentioned being an animal lover or being opposed to animal cruelty. This intervention is less plausible without significant regulatory pressure or voluntary commitments from AI developers, as it would, if not sufficiently subtle, be contrary to commercial incentives.

Finally, AI could allow for expert knowledge about animals, their welfare, and their cognitive abilities to become more accessible to the public, which would close the gap between popular opinion and deeper scientific understanding. There are countless studies showing that animals have abilities and cognitive capacity beyond what many thought possible. A study from 2009 found that pigs had the ability to interpret a mirror image in order to find food. Koalas appear to be capable of predicting what will happen in a particular situation based upon experiences from their past. Families of dolphins have been known to change the location and time of their hunt to avoid human activities. The results of these studies have been shared in the media, but these findings could be disseminated in a more targeted and robust way when AI makes access to this information easier. The information could also be found by asking LLMs, which is often easier for people than searching for studies themselves. When people know more about what animals are capable of and how rich their inner lives might be, this could prompt the acceleration of moral circle expansion.

Economic transformation

Moral progress rarely occurs in isolation from material conditions. Social change, even when moral circle expansion is almost certainly a factor, is also often partly triggered by economic shifts. The abolition of slavery in the late 18^th and early 19^th centuries in Britain coincided with the Industrial Revolution; researchers have identified a strong relationship between the economic change connected to an evolution in manufacturing processes and support for the abolition of slavery. Industrialists did not rely as much on income from slavery and their purported values reflected that material reality. With that in mind, it is likely that a significant economic shift triggered by the development of increasingly advanced AI would have an impact on how people feel about animal exploitation. How could AI change the economy in a way that could benefit animals?

Cost or income level has consistently been shown to be a significant factor in determining food choices. When asked, consumers said that price was among the top three considerations when it comes to whether or not they would choose plant-based protein products. Thus, affordability is a crucial consideration when we think about how meat products compare to alternatives, and how to make the latter more appealing or accessible to consumers. Nielsen data ‘demonstrates that, on average, plant-based meat is 2x as expensive as beef, more than 4x as expensive as chicken, and more than 3x as expensive as pork per pound’. AI-driven research and development could accelerate alternative proteins towards price parity. As dominant AI-enabled designs emerge in alternative protein development, costs in this area will decrease. I will discuss alternative proteins in more detail in the section on farmed animals, so I will not discuss in depth here.

AI-enabled automation could reduce the labour cost advantage of animal farming. As it stands, the animal agriculture industry relies heavily on cheap human labour; slaughterhouse workers, factory farm operations workers, and farm hands are generally low-paid, and this low cost of labour is a factor in making animal products deceptively inexpensive to the consumer. If much of the labour involved in both animal agriculture and alternative protein production is automated, the former loses some of its cheap labour advantage. Automation levels the playing field: animal agriculture loses a cost advantage it currently holds, while alternative proteins shed costs they were already bearing. AI could enable an accurate cost-analysis of external factors in animal agriculture. The true costs of the industry, including environmental degradation, contribution to climate change, antibiotic resistance, and public health risks from zoonotic diseases, are substantial but largely hidden from consumers and investors. One estimate puts the annual hidden costs of the global food system at over $10 trillion, with animal agriculture a significant contributor. AI could make these costs legible, undermining the perception that animal products are cheap to produce.

AI could make animal welfare issues legible to capital markets in a way they weren't previously. Through automated monitoring and reporting of welfare data, AI could make animal welfare visible to investors, potentially negatively impacting the attractiveness of animal agriculture as an area for investment. FAIRR's report ‘Factory Farming: Assessing Investment Risks’ identifies at least 28 environmental, social and governance (ESG) issues that have the potential to damage the financial performance of factory farms. Investment risk frameworks are beginning to treat factory farming the way they treat fossil fuels, as an industry with significant stranded asset risk from regulatory change, pandemic exposure, and reputational damage. One example they mention is California's gestation crate ban (supported by over 60% of voters), drawing attention to this as a regulatory shift that investors should be pricing in. The report concludes that ‘there are signs that the financial community is beginning to recognise the need to ensure that investments in agriculture, including meat production, are responsible and do not negatively affect people, livelihoods and the natural environment’. 

With AI tools increasing transparency around animal welfare practices, the animal agriculture industry could be negatively impacted by increased hesitancy from investors. For example, AI monitoring could report an animal welfare or health violation on a farm or in a slaughterhouse, taking away from the marketability of the associated products. Additionally, since the COVID 19 pandemic, there is an increased awareness around the possibility of further zoonotic disease pandemics, which presents a further financial risk to consider when it comes to animal farming. AI facilitates transparency around welfare, disease prevalence, and other factors, and it is plausible that this transparency will damage the animal agriculture industry. 

Epistemic Shifts

For centuries, the dominant Western view held that animals were mere machines. In the 17^th century, Descartes wrote about animals as ‘automata’; Malebranche summed up Descartes’s view when he wrote that animals ‘eat without pleasure, cry without pain… desire nothing, fear nothing, know nothing.’ Later, Darwin’s theory of evolutionary continuity argued that species of animals, including humans, have differences that are of degree rather than of kind. We are all made up of the same fundamental elements, but they present to varying degrees of significance across species. Primatologist Frans de Waal, who wrote Are We Smart Enough to Know How Smart Animals Are? conceived of animal cognition in a similar way; he argued in his book that, rather than thinking of intelligence as a ladder, with ourselves at the top and other species on various lower rungs, it would be more helpful to think of it as a bush, with different kinds of cognition appearing in ways that are, in their alienness, incomparable to our own. As technological development races ahead, how might AI help us to better understand animal minds?

To start with, AI could help in significant ways when it comes to research on animal cognition and behaviour. Machine learning could analyse patterns of behaviour at an enormous scale, looking at monitoring data, thousands of hours of camera footage or microphone recordings, and study outcomes and extracting relevant findings and conclusions. AI could identify subtle social dynamics or problem-solving behaviours at a speed and scale that would be difficult for humans without substantial investment of time, energy, and money. It could also eliminate much human error that can occur over so many hours of monitoring, leading to more accurate, and therefore useful, results. Automated recording and testing could also be conducted by AI, allowing for much larger sets of data to be collected than cases in which monitoring would need substantial human oversight.

Neuroimaging can be, and is, used on animals as well as humans to monitor brain activity. For example, functional MRI is used on rodents and can be useful for observing patterns of neuronal activity across the entire brain. AI is already being used to assist in neuroimaging for humans, but many of the advantages are relevant for animal studies too. AI can increase the speed of data processing, offer faster and more accurate interpretation of images, and reduce errors. Less time and labour needed for research means that more research can be done within the same timeframe or budget, allowing animal cognition research to accelerate. AI also has the ability to read MRI scans with ‘reduced or without contrast without significant loss in sensitivity for detecting lesions’. This matters for animal welfare as gadolinium-based contrast can cause uncomfortable or even painful reactions, posing a serious danger to health and life in rare instances. The ability of AI to efficiently and speedily manage large sets of data is also important when we look at how it could be used to compare one animal species to another. AI could identify differences and similarities across species, including in brain scans where possible, which could hint at shared cognitive capacities. That might be especially impactful when it comes to an epistemic shift if AI identified more similarities between humans and non-human animals than we previously imagined there would be.

Another potential AI-enabled trigger for an epistemic shift is advances in interspecies communication technology. A number of organisations have made significant progress towards the goal of understanding and communicating with other species. Project CETI, a non-profit founded in 2020, deploys machine learning, robotics, and large-scale acoustic datasets to record and interpret sperm whale communication over long distances, and they have uncovered a 'sperm whale phonetic alphabet' by linking behavioural patterns to vocalisations. Earth Species Project (ESP), founded in 2017, also uses AI technology to enhance interspecies understanding. Their team of technology and non-profit leaders, AI researchers, engineers, and ethologists leverages advances in large-scale models, multimodal learning, and extensive datasets to allow for greater understanding between humans and non-human animals, including not only whales, but crows, elephants, orangutans, and others. Interspecies Internet is a think-tank that works to accelerate interspecies communication; the project's premise is that it could be used to 'link non-human species that are not collocated and leverage computational capacity to support the use of AI/ML methods in transducing signals from one species into coherent signals for another.' It is plausible that advances in AI technology will allow us to better understand and communicate with non-human animals, triggering an epistemic shift in how we think about animal minds.

We have seen recently how more research and information on animal cognition can cause this kind of shift with the substantial opposition to octopus farming. In 2019, Nueva Pescanova, a Spanish seafood company announced intentions to open the world’s first octopus factory farm in Gran Canaria with the aim of producing one million farmed octopuses annually; they intend to begin operations in 2027 but have not yet secured the necessary approvals. Several MEPs at the EU level opposed the farm, and sent a letter to the government of the Canary Islands. Researchers from the Animal-Human Policy Center surveyed 14,131 adults across 13 EU countries and the UK, and found that about half of respondents supported national bans on intensive octopus farming, and fewer than one in five opposed such bans; they went on to say that educational messages about ‘octopus sentience and welfare were the most persuasive overall’. Additionally, in 2024, California became the second US state to ban octopus farming, with legislation explicitly referring to the animals’ high level of intelligence[2]. This widespread resistance to octopus farming has been fuelled in part by popular works like Peter Godfrey-Smith's Other Minds and the Oscar-winning documentary My Octopus Teacher, which brought octopus cognition to mainstream attention. AI-enabled research could accelerate these kinds of epistemic shifts across many more species, making it harder to maintain ignorance about the animals we farm at scale.

Digital Minds

It is possible that we will create digital minds – conscious AI systems whose moral status we will then be forced to grapple with. In this brief section, I will explore two possibilities related to the emergence of digital minds; the first is the knock-on effect that our moral consideration of digital beings could have on animals, and the second is the idea that agentic digital minds could have values that are different – perhaps in some ways better – than our own.

A number of the statistics included here were gathered and utilised in the ‘Moral status of digital minds’ problem profile on 80,000 Hours. The 2020 PhilPapers Survey asked 1785 English-speaking philosophers from around the world the following question: ‘Other minds (for which groups are some members conscious?)’. 3.39% said that they accept or lean towards current AI systems being conscious, but 39.19% said the same about future AI systems. Another survey was distributed on two occasions of the annual meeting of the Association of the Scientific Study of Consciousness – in 2018 and 2019 – to 166 consciousness researchers from backgrounds including philosophy, neuroscience, psychology, and computer science. When asked the question ‘At present or in the future, could machines (e.g. robots) have consciousness?,’ 67% answered either ‘probably yes’ or ‘definitely yes’. Philosopher David Chalmers focuses much of his work on the hard problem of consciousness and the possibility of AI consciousness, considering the terms consciousness and sentience as equivalent in this context. He has stated that he believes there is a roughly 25% chance that we will have conscious AI systems in the next decade, though he also argues that there is at least a one-in-three chance that biology is required for consciousness. Arguments about the likelihood and timing of digital consciousness are extensive, so I won't delve deeper here. With that being said, the possibility of the emergence of digital minds in the coming years is non-negligible, and therefore it is plausible that we will, in the not too distant future, be forced to reconsider what is necessary for moral patienthood.

Digital minds and animals are often considered together, categorised under the umbrella of ‘non-human sentient beings’. In 2025, Sentient Futures organised a conference on AI, Animals, & Digital Minds (AIADM), which aimed to ‘bring together leaders in AI and non-human ethics to share ideas, resources and opportunities,’ and to collaborate on the goals of using AI to help animals, and ensuring that AI goes well for all sentient beings into the future. The organisation explicitly states that its aim is to reduce suffering for all sentient beings – both animals and digital minds would be included here. The emergence of digital consciousness, if it does happen, would force people to consider whether moral patienthood can meaningfully apply to agents that are not entirely human-like; the degree to which a digital consciousness would resemble a person is unclear, although it is unlikely that they would be embodied in any recognisably human-like way. If people were able to expand their conception of which beings are worthy of moral consideration to include those not like ourselves, it is possible that this could open the door to other kinds of moral circle expansion. To illustrate this point, I will refer to an article by Brian Tomasik in which he writes that it was ‘only because I already cared about animal welfare that I even considered the idea that powerless subprocesses of a future intelligent civilization’s computations might have moral standing.’ Could this work the opposite way? If people cared about digital minds, could they come to care more about animals too? This outcome is possible.

Secondly, and more speculatively, if digital minds did emerge, it is possible that they would have morals and values that deviate from our own. Would they be more convinced by the arguments of animal activists and vegans, and want to act in the interests of animals? It sounds like a strange and unlikely prospect, but perhaps we cannot discount it. We have reason to believe that, even if digital sentience did emerge, it would take some time for us to recognise this emergence. During this time of ignorance, we would almost certainly continue using digital persons as we please with no regard for their wellbeing. Once we did acknowledge them, if we did, they would already have experienced oppression by humans. Research suggests that there are higher rates of veganism among historically oppressed groups such as black people and women than among the general population. The reasons are likely multi-factorial, but one plausible explanation is a connection between the experience of being oppressed and a broader commitment to anti-oppression. Digital persons could, therefore, be more likely to side with the oppressed. In the case of AGI, or even more so with ASI, the emergence of agentic digital persons could significantly limit the ways in which AI could be used to harm animals and, simultaneously, accelerate AI use in improving their status and welfare.

These foundational shifts could reshape the landscape for animals across every domain. The most immediate and large-scale impact would be seen where we currently cause the most extreme levels of animal suffering: farming.

Farmed Animals

The vast majority of the suffering and death that we inflict on animals is done through food systems. There are over 20 billion animals on factory farms at any given time, with many more being farmed in other environments or being wild-caught. The animals we kill for food can be broadly broken down into land animals (such as cows, chickens, pigs, sheep, goats etc.) and aquatic animals (including fish, octopuses, crabs, shrimps etc.). The former are much easier to find accurate numbers on, as aquatic animals caught or farmed are often measured in terms of weight rather than individual animal numbers. The situation for farmed animals, at the moment, is grim. However, if we manage to make AI go well, there are many ways that this could turn around. There are two possible positive futures: one in which we massively improve the conditions for farmed animals, and one in which we replace animal farming altogether. Deciding whether to focus on just one of these paths or both simultaneously echoes debates about welfarism versus abolitionism - I won't get into that here, but this post is helpful for considering this conflict (if there is one). 

Improving Conditions

Critics have understandably raised concerns that precision livestock farming (PLF) could be used to intensify farming rather than improve farmed animal welfare, but here I will focus on its potential to reduce suffering. Precision livestock farming can include technology such as health monitors, cameras, audio recording technology, feeding sensors, and more with the aim of managing ‘individual animals by continuous, automated, and real-time monitoring of health, welfare, production/reproduction, and environmental impact’. To examine how PLF could help improve animal welfare and reduce the number of on-farm deaths, it is worth first looking at what causes these deaths. Although PLF might contribute even more meaningfully to reducing suffering than reducing on-farm mortality, it is substantially more difficult to find data on suffering; thus, I will use causes of on-farm mortality as a proxy for causes of on-farm suffering too – it is probable that the causes are roughly equivalent. Across numerous studies[3] and species of land animal, the most common causes reported were: 

1. Disease (such as arthritis and peritonitis) 

2. Metabolic and digestive disorders

3. Accident or trauma (including crushing, drowning, electrocution)

4. Inadequate (cramped, cold, or dangerous) living conditions

PLF is already being used on some farms; as stated previously, the industry is already worth billions of dollars. However, AI advances would enable the further development of this technology, almost certainly improving its capacity, efficiency, affordability, and accuracy. There are too many farmed animal species to discuss PLF application to each in detail, but I can offer examples of how widespread use of improved PLF technology could drastically reduce suffering on farms. Peritonitis, as well as many other similar diseases, can trigger reduced appetite, reluctance to move, weight loss, and a swollen abdomen – technology that monitors the feeding behaviours, weight, and movement frequency of individual animals could identify these diseases. Metabolic and digestive issues can be identified in animals by monitoring feeding behaviours, waste/digestive efficiency, and visible distention. Feeding sensors and surveillance technology could pick up on each of these factors, flagging abnormalities early and allowing issues to be dealt with before they become debilitating or deadly. AI-optimised feed formulations could help to reduce painful conditions, such as leg deformities in broiler chickens from rapid weight gain, that are directly caused by feed designed for maximum growth rather than wellbeing. Surveillance technology and stress monitors could sound the alarm if an animal was at risk of drowning, was trapped and at risk of being crushed, or was caught in a fire. Finally, by closely monitoring the health data of individual animals on farms, it would be easier to determine whether living conditions were adequate – behaviour that indicated high levels of stress or low mood could be recorded and changes made accordingly. 

A similar model could change the lives of countless animals in aquaculture. Aquaculture is the fastest-growing food production sector globally. At any given time, there are 125 billion fishes on farms and 440 billion shrimps are farmed each year. There are other animals to consider too, such as crabs and lobsters. In terms of numbers alone, the potential impact of improving animal welfare of these farms is massive. Some of the main welfare concerns for animals in aquaculture include:

1. Poor water quality

2. Overcrowding

3. Disease

4. Handling and transport

Precision aquaculture (PA) could be automated to monitor the water quality accurately and consistently, checking the pH level, concentration of ammonia, temperature, and salinity, and sounding the alarm when water quality dips below a certain level. Surveillance technology and feeding sensors could look for signs of stress in animals, such as reduced appetite, erratic movements, increased aggression, and repetitive behaviours. Signs of disease, such as ulcers and cloudy eyes, could be identified by cameras and flagged as concerning, allowing affected animals to be treated or removed to protect the rest of the animals from infection. 

Conditions could be improved on farms, but what about in slaughterhouses? Although animals will always be killed there – that is, after all, their purpose – there is substantial room for improvement when it comes to suffering in slaughterhouses, and AI could help us in this respect too. A number of the biggest welfare concerns in slaughterhouses are ineffective stunning, pre-slaughter stress, and pressure on workers to maximise efficiency. Widespread implementation of advanced AI monitoring in slaughterhouses could allow ineffective stunning to be identified and flagged to workers before animals move on to the next part of the slaughter process, ensuring that animal stunning is consistent. AI could analyse live video and audio of animals being led from the trucks to the slaughter line, identifying signs of distress (stressed vocalisations, backing away) or danger (crushing, slipping, pile-ups). The slaughter line speed could also be optimised for welfare and efficiency rather than efficiency alone – this is especially relevant for animals such as chickens who are literally shackled to a conveyer belt system. If the slaughter of an animal is slightly slower than average, the belt could automatically adjust, allowing all animals to be killed in a way that causes the least suffering.

Finally, campaigns to improve welfare legislation for farmed animals could be substantially more effective with AI monitoring tracking farmed animal wellbeing. There would be more data than ever before to refer to when pushing for policy changes, and compliance to welfare standards would be easier to audit – the footage would be available to auditors. This would also mean that any purposeful cruelty would be easier to spot quickly, record, and prosecute (such as in 2024 when secret cameras at Midland Bacon at Carag Carag captured footage of a man allegedly sexually abusing a pig at the facility). In the future, PLF could mean that farms and slaughterhouses are full of cameras, none of which are secret, which would almost certainly reduce the risk of acts of blatant animal cruelty.

Replacing Animal Agriculture

What if, instead of (or as well as) improving the conditions of farmed animals, AI allowed us to make animal farming obsolete altogether? That is the dream of many working in alternative protein development. Alternative proteins, which can be made from plants, or via cultivation or fermentation, are designed to be consumed in place of meat – using less land and water, creating less pollution, and causing far less suffering. The Good Food Institute, a non-profit think tank focused on innovations in alternative proteins, claims that this kind of research can help to ‘produce food that people love and usher in a more sustainable, secure, and just food future’.

Plant-based proteins are usually primarily derived from soy, peas, wheat, rice, or mung beans. They can, however, also come from microalgae or fungus (in the case of mycoprotein, which comes from fusarium venenatum). Cultivated meat, also known as lab-grown meat or, more recently, clean meat, is made by growing animal cells in controlled conditions, rather than farming animals in a traditional way. Advances in AI could accelerate innovation and research in the area of alternative proteins in a number of ways. Max Taylor has covered this topic extensively in his article, so I will only briefly outline some key points here. In order to consider how advanced AI could accelerate progress in the alternative protein space, it is important to first think about some of the steps involved in creating plant-based or cultivated meat. I will explain the processes quickly and simply, as that is sufficient to make my points.

Let’s start with plant-based meat. Initially, crops must be grown and harvested. Protein is extracted from the plants in a lab, separating it from the other parts of the plant such as fibre or starch. Plant-based fats such as coconut are added, and binders are used to hold everything together. Flavourings and colour are added, allowing companies to replicate the taste and appearance of animal-based meats. Extrusion is the next step, where the mixture is exposed to high heat and pressure – this step is important for mimicking texture. Finally, the product is shaped into whatever it is meant to be (burgers, nuggets, sausages etc.). Cultivated meat involves taking cells from a live animal, putting them in a bioreactor with a growth medium, giving the cells time to replicate, giving them a scaffold so they replicate the structure of whatever tissue you’re replicating, harvesting the tissue, and processing into food. How could advanced AI help with any of those processes?

Firstly, AI could optimise crop management through the collection and monitoring of real-time sensor data and satellite imagery of the crops required, utilising information about growth rate, signs of disease, and more to make recommendations to farmers to help maximise yield. In the case of cultivated meat, AI could analyse which individual animals or animal breeds have cells that proliferate best in culture, ultimately reducing the number of biopsies needed. It could be used to  optimise the parameters in plant-based protein development such as moisture content, nutrient profile, growth factors, and temperature, making the fermentation or extrusion processes better. The same could be said for cultivated meat, where AI could optimise formulation of growth medium. By analysing millions of natural proteins, it could quickly identify which ingredients meet specific functional needs and predict how different ingredients and molecules will behave before they are put into production in a lab environment. By enabling more efficient research and production, AI could make manufacturing more efficient overall, speeding up the production of better and better alternative protein options. The automation of many of the processes could make them cheaper, thus allowing alternative proteins to reach price parity with animal-based meats, making them more appealing to the consumer and triggering potentially rapid market growth. Finally, AI could be used in marketing efforts by alternative protein companies, again saving these companies money and allowing them to target the most amenable potential customer bases.

If alternative proteins were equal to or better than animal proteins in terms of taste, texture, price, and availability, then it is possible that farming animals for food could eventually become obsolete. This future may not be probable, but it is not impossible and, with the help of advanced AI technology, we may someday be able to make it happen. It’s a big dream, but it has legs.

Wild Animals

Although the present scale of suffering caused directly by humans is highest in animal agriculture, the scale of animal suffering overall is highest in the wild animal population. For every human on earth today, there are roughly 10-50 wild birds, 10-100 wild mammals, thousands of wild reptiles and amphibians, millions of insects, and as many as 100,000 fish. Given those numbers, the cause of reducing wild animal suffering seems incredibly important. However, due to tractability concerns, many can be hesitant to support wild animal welfare interventions; some are concerned that nature is too unpredictable and that interventions could cause more harm than good. Here, we will look at some of the means through which AI could help us to reduce wild animal suffering and death and, in some cases, in ways that are more tractable than many current interventions.

Starvation & Thirst

Starvation and thirst are some of the most significant threats to wild animal welfare globally. Drought is a major contributing factor to food and water availability. Researchers at the AI for Drought project aim to improve the accuracy of drought predictions using AI technologies, and a 2024 study found that AI models that were trained on climatic datasets outperformed conventional drought indicators in prediction capabilities. AI-enabled accurate drought predictions and ecosystem monitoring for crop failures would allow us to identify areas that are high-risk for near-future mass animal deaths. Pre-empting high-risk areas for such events would enable early intervention such as digging waterholes or installing guzzlers (long-term solutions) or providing emergency water (bringing large water containers to affected areas). 

As well as extreme weather conditions, starvation can be the result of overpopulation in an area, leading to competition for scarce resources. Overpopulation is often addressed through culls, as is the case in Ireland where approximately 78,000 deer were killed in 2023. An alternative that avoids a cull and prevents starvation would be investing in the development of less expensive and more easily administered forms of contraception. AI could accelerate research and development in this area by, for example, modelling contraceptive compound efficacy to enable more effective administration, or optimising delivery mechanisms. Contraceptive darting for deer can be effective, but is only realistic on a small scale currently due to the high resource investment required. Enhanced drone and tracking technology could also assist in this effort, allowing deer populations to be monitored more closely, and darted animals to be more easily identified. Oral contraceptives administered through feeders are more cost effective, but currently in development and only available for a small number of animal species such as rats. 

Natural Disasters

As with droughts in the previous section, AI could improve our ability to accurately predict extreme weather events. With access to large enough good quality datasets, AI can predict natural disasters such as earthquakes, hurricanes, and floods, analysing seismic data, rainfall records, and satellite images. Two limitations with current AI technology are that it is prone to errors in predictions, and it can only base predictions on past natural disaster records so it is limited when it comes to processing changing trends. More advanced models could improve accuracy, making predictions more actionable. If it were possible to predict, with reasonable accuracy, when a flood or hurricane were coming, it would be possible to pre-install wildlife deterrents in high-risk zones, activated when a disaster is imminent. AI-powered animal deterrent systems can use loud noises and flashing lights to ward away nearby wildlife, thereby potentially reducing animal suffering and death from natural disasters.

AI technology could also allow us to help wildlife in the aftermath of a natural disaster. After a flood, for example, many land animals can be stranded, in trees or stuck on trapped areas of higher ground. Thermal imaging and drones might allow us to find animals after a flood, landslide, or earthquake, and rescue or relocate scared and injured animals where possible.

Conservation

AI could assist with wildlife conservation efforts. There are already a number of AI applications in the field of conservation, but more advanced AI models would have capabilities beyond what we currently see. The Conservation AI Network has developed AI models that can analyse footage from drones and camera traps to identify wildlife and track movements. They also use this model to analyse footage submitted through their online platform, with users receiving a notification if an animal of interest is detected on their submitted footage. Surveillance technology of this kind could become more advanced, capable of detecting endangered species from a distance or from sounds undetectable to the human ear using live cameras and audio recordings – this would be an incredibly useful tool for conservation efforts. Such technology would further assist in anti-poaching efforts. Acoustic sensors that detect animal calls could similarly be used to identify gunshots in wildlife protected areas and motion sensor camera traps could detect human presence where there should be none; both of these are current applications of AI technology, but they could become more widely used, cost effective, and accurate in the future, allowing us to save even more animals from death at the hands of poachers and prosecute these wildlife crimes through identification from camera trap footage.

Better AI-enabled tracking and identification of specific animal families and communities could allow us to place previously injured and rehabilitated animals back with their original groups, improving the likelihood that they will go on to thrive. Finally, AI could, as with PLF, allow for better welfare of animals in rescue centres, sanctuaries, and breeding programs by monitoring relevant health, behaviour, and welfare metrics and optimising care accordingly.

Road Deaths

Every year, millions of large wild animals such as deer, moose, and kangaroos, are injured and killed globally on our roads. One potential solution that could drastically minimise those numbers is AI-enabled vehicle detection systems that can detect large animals that are approaching. Studies have been conducted on this type of intervention as far back as 2009, and yet systems of this kind have not undergone rigorous testing or evaluation and are not deployed at scale. AI has since enabled rapid improvements in detection technology, meaning modern systems would be substantially more effective than those discussed in 2009. In the future, the technology could be better again, detecting more species of animals from a further distance and with more accuracy. AI could optimise this technology and, in doing so, make it more feasible to require, by legislative measures, this technology in all new cars.

AI cameras, thermal imaging, or animal tracking could help to identify high-wildlife-traffic areas even at night or in areas of substantial forest cover where we could install animal deterrent systems or wildlife crossing bridges beside and across roads. Deterrents could sound alarms when cars are coming, potentially frightening away animals that would otherwise cross – for cases in which this was ineffective, the aforementioned animal detection system in the car could be sufficiently advanced to enable the avoidance of a collision even with faster animals like deer and kangaroos. Finally, self-driving vehicles, with their built-in detection systems, faster reaction times, and lower error rates than human drivers, would likely hit fewer animals. A two-year survey conducted in the US found that driver error accounted for over 90% of road collisions – driver error is not a relevant factor with autonomous vehicles. There is no reason we couldn't drastically reduce road deaths in the future.

The interventions that I have discussed here primarily address anthropogenic harms. However, many threats, including predation, disease, and parasites, to wild animal welfare are not caused by us and remain somewhat beyond our current ability to address effectively on a large scale; crucially, our predictions regarding the outcomes of possible interventions are subject to substantial error. In the 1960s and 70s, we introduced Asian carp, now often referred to as invasive carp, to the US to control algae overgrowth; now, the US spends approximately $70 million annually trying to control carp overpopulation, with many millions of fish killed every year in this effort. AI-enabled ecosystem modelling could help address uncertainty and prevent these kinds of problems by allowing us to simulate the effects of these kinds of interventions before they are implemented, testing questions like whether suppressing a parasite leads to overpopulation, or whether feeding draws predators to an area. This kind of simulation would make wild animal welfare interventions more tractable, potentially unlocking a largely neglected cause area with enormous scale. 

Companion Animals

When we consider the welfare of companion animals, we can divide them into two groups: companion animals with pet owners[4] and those without. The welfare issues will be somewhat different between groups; in the former, welfare can be improved with optimised tools for providing adequate care such as diagnostic tools, and in the latter, some of the biggest improvements in welfare would come through means of providing homes or owners. Although it can be tricky to find comprehensive data on pet ownership across all species, approximately one in three households globally has a dog which hints at the scale involved. At the same time, in the US alone, over 600,000 unwanted animals were euthanised in shelters. Due to scale, reducing the suffering of companion animals, both housed and unhoused, would be impactful – in the future, AI could help us to do that.

Veterinary Diagnostics

When it comes to veterinary care, AI could make diagnoses faster, more accurate, and cheaper, and streamline administrative tasks. Vetscan Imagyst, for example, which is described as a ‘veterinary AI analyser’ offers expert-level analysis across multiple tests. These include blood smear analysis, dermatology diagnostics (identifying yeast, bacteria, and inflammatory cells), and fecal analysis for parasites. Traditional diagnostic tests often require laboratory analysis, whereas AI could allow for in-clinic analysis, cutting down the wait time to diagnosis substantially, and eliminating transport and lab costs. AI can also be used across imaging modalities such as radiology, ultrasound, and MRI to identify abnormalities. These applications could help to weaken two barriers to high standards of veterinary care: wait times to diagnosis and cost. Early diagnosis leads to better outcomes, and point-of-care diagnosis using AI technology could enable this. 

The American Veterinary Medical Association conducted a 2023 Pet Owner Attitude Survey which found that affordability was the second highest priority when it came to veterinary care decisions, after pet health and safety. Many owners struggle to pay vet bills, and a 2024 petition which called for ‘regulation to ensure fairer and more transparent vets bills’ was signed by more than 80,000 people. By making diagnoses faster and, ultimately, less labour and resource intensive, and by eliminating many administrative tasks through automation, veterinary care could become cheaper and more accessible, improving pet care standards.

Health & Disease

Due to advances in technology, DNA testing for pets has transformed from a novelty – in which people could learn their dog’s breed mix – into a tool to screen for genetic predispositions to specific diseases, allowing carers to manage their pet’s health more effectively. In recent years, researchers have identified hundreds of genetic variants that can lead to health problems in pets. By identifying these genetic variations in your pet, you could make informed early decisions on medication (some dogs have genetic variations that increase sensitivity to medications), behavioural interventions (to manage fear, anxiety, or aggression), and breeding. Wisdom Panel is just one example of a company offering AI-enabled genetic testing for pets. 

Beyond genetic screening, AI also enables real-time health monitoring. Health monitoring for companion animals could mirror aspects of precision livestock farming, involving tools such as feeding bowls with sensors, litter trays with sensors, surveillance technology, and wearables (such as collars) that can monitor stress levels. AI-powered automatic feeders for pets could adjust the food amount and frequency depending on pet needs and behaviour, optimising nutrition. Apps such as TTCare can analyse companion animal body language to judge wellbeing, and apps such as Daisy allow pet owners, some of whom may not find traditional veterinary care accessible, to ask personalised advice about their pet’s health. Apps such as these could become more advanced, providing more accurate screening and health assessment, further improving outcomes for companion animals.

Interspecies Communication

AI-enabled communication with, and understanding of, animals is a huge topic that includes many different animal species, both wild and domestic. I have written at length about some of the applications and ethical implications of this technology, so I won’t delve into too much detail here. However, one of the fastest growing areas of interspecies communication is with companion animals. Research carried out by scientists at the University of California San Diego found that dogs can recognise and respond appropriately to verbal cues; for example, when a button that vocalises ‘outside’ is pressed, some dogs looked towards the door. The Jeremy Coller Centre for Animal Sentience began work in 2025 on research exploring animal consciousness, with one of its projects focusing on how AI might enable humans to communicate with their pets. Furthermore, the Coller Dolittle Challenge, started in 2025, rewards interspecies communication innovation with an annual $100,000 prize.

Baidu, a Chinese multinational tech company, filed a patent in 2025 with China National Intellectual Property Administration for a system that aims to convert animal vocalisations into human language, according to the patent document. The system would collect data about your pet, including behavioural patterns, vocalisations, and physiological signs of mental states to allow us to better understand our pets and to foster ‘deeper emotional communication’. What does this mean for pet wellbeing? If AI allowed us to understand our pets with a high degree of accuracy, we could identify sources of distress or discomfort early, optimise living conditions, be notified more easily of injury, soothe fear and anxiety, and more. This application of AI technology has a more subtle impact on wellbeing than others mentioned previously, but it is also probable that we do not yet know how this technology would change our relationships with our pets; it may have a greater impact than we can foresee. 

Rehoming & Matching

As mentioned in the introduction to this section, the number of unwanted animals euthanised in shelters is huge; globally, millions of animals are killed every year. A report compiled by Hill’s Pet Nutrition (using findings from a YouGov survey of 1,505 US adults and interviews from welfare professionals) found that the following were barriers to adopting animals from shelters:

1. Housing policy restrictions

2. Unable to find desired companion in shelter

3. Financial constraints and cost of veterinary care

Let’s consider how advanced AI technology might be able to address some of these barriers to adoption. When it comes to housing restrictions, the potential for AI applications to help is limited. Apps could be created and optimised to allow pet owners to find pet-friendly accommodation, but beyond that, the biggest changes here would require changes in policy and renting norms for which AI can help only minimally or indirectly (for example, through administrative tasks). As discussed in a previous section, AI could facilitate the availability of less expensive veterinary care, and apps that function in place of vet visits for minor issues. Perhaps the biggest effect that AI could have in this space would be through optimising matching for pet seekers and animals in shelters. AI-powered adoption platforms such as Petfinder and WeRescue could improve compatibility analysis, ensuring that matches are optimised for long-term success and bringing down the return rate in shelters. Users of shelter platforms could set up alerts for the kind of pets they are looking for, including specific breeds – shelter uploads could be automatically analysed to identify breeds, alerting relevant users when matches appear (potentially even identifying cases where the desired breed is likely to be part of the mix). Finally, on a different but related point, AI tools could be used to scan ads online and identify red flags in pet listings such as frequent ads for puppies from the same user, multiple breeds available at once, and puppies sold before they’re old enough, drawing attention to potential puppy mills.

Welfare of Strays

Stray animals, mostly dogs and cats, can present threats to both human and nonhuman animal welfare in areas where populations are high, such as Morocco and India. For example, approximately 100,000 people are bitten by one of Morocco’s 3 million stray dogs per year, with 40% of them under 15 years of age. These dogs also present a disease risk from rabies, leishmaniasis, and echinococcosis. Controversially, in advance of the 2030 FIFA World Cup, there have been stray dog culls in some cities in Morocco, with one young woman remarking that ‘walking to school, [she] would pass pools of blood on the street… I realized it wasn’t normal to start your day stepping over dead bodies.’ In India, where the stray dog population hovers around 62 million, there have been attempts to manage the population through trap-neuter-release programs but the numbers make this an enormous task. How could AI help us to manage the welfare of stray animals to the benefit not only of the animals themselves, but to the people that live in these high-density areas?

AI-powered image recognition could identify individual dogs from CCTV or drone footage, allowing specific dogs to be identified and signs of illness or injury to be spotted (this could be especially useful in areas where rabies poses a serious threat to public health). By analysing CCTV footage, shelter intake data, and sightings logged online or on rescue apps, AI could allow us to identify areas where strays are living in highest numbers, allowing us to optimise the efficiency of vaccination campaigns, feeding efforts, or sterilisation drives. Being able to identify individual animals also means that we can monitor which animals have been vaccinated or sterilised and which haven't, allowing for further targeting of relevant populations. Oral contraceptives could also be an option to explore here in the future if AI accelerates development in this area, helping to humanely control numbers. Finally, by improving technology that can detect individual animals, we would allow lost pets to be reunited with owners – or for those who abandon pets to be held accountable.

Other Animals

There are many ways that we use animals beyond what is detailed above. We use animals in fashion (leather, fur, wool, silk), in entertainment (zoos, aquariums, circuses, in movies and TV shows), for testing medical and cosmetic products, for transport (horses, donkeys etc.), and in security and service roles (guide dogs, drug-sniffing dogs, mine-detecting rats etc.). In this brief section, I will look at a few of these and how AI technology could improve the future for animals.

Vivisection

Animals, most commonly mice and rats, are used for testing drugs, medical supplies, cosmetics and toiletries, and even household cleaning products. AI could help reduce the amount of suffering and death from vivisection in numerous ways. Firstly, it could analyse vast amounts of data on animal tests done previously, calling attention to relevant results and identifying links that humans might miss – this could negate the need for some future tests. New AI systems are beginning to be able to measure the toxicity of new chemicals; given that many animal tests are carried out to test exactly this, a future AI that could consistently and accurately do this would end tests for this purpose. AI-enabled research on organoids and cell cultures could allow for the testing of new drugs without animals. 'Organ chips' (‘clusters of cells embedded in diminutive electronic devices that simulate an organ's behaviour’) are becoming increasingly sophisticated, and increasingly complex organoids are being created that more closely simulate biological reactions, with some even having the ability to circulate blood. Predictive modelling of disease progression could help to eliminate the need for longitudinal animal studies and AI’s analysis of brain organoids could offer insights into human brain function without the need for neurobiological animal tests.

Fashion

Many of the same AI applications discussed in previous sections apply here. AI could help us to move away from animal-based textiles such as leather and wool, and towards more animal-friendly and sustainable alternatives. Innovative vegan leathers can be made from pineapple leaves, mushrooms, or apple peels. AI could help, as in the case of farming, to make crop management as efficient and cost-effective as possible. It could be used to compare vegan materials to animal materials in terms of texture and appearance, and help make decisions about the best materials to use; with a view to this, it could also analyse supply chains and determine which materials are the most sustainable and cause the least environmental damage, indirectly benefitting wildlife. In production, AI could automate many of the processes to avoid human error and minimise the number of garments that need to be discarded or remade – cutting down on waste, water usage, and costs. Lower costs would mean that vegan fabrics would become more competitive in the fashion market and therefore more attractive to designers and fashion companies. Effective AI-enabled marketing campaigns could help to identify target markets and drive demand away from animal-based textiles. By cutting down on, or eliminating, animal materials such as leather, wool, down, fur, and silk, many animals could be saved from lives of suffering and premature death. The suffering is especially acute when it comes to fur, as animals on fur farms regularly suffer appalling treatment, with some skinned alive without stunning or anaesthesia. AI might help us to empty these cages.

Entertainment

Although the level of animal welfare in zoos varies substantially, there have certainly been many cases of animals kept in enclosures that are too small, without adequate food, and with no forms of stimulation. The welfare concerns with zoos are substantial enough to have created a psychological disorder in animals referred to as ‘zoochosis’, in which animals engage in obsessive, repetitive behaviours such as pacing, swaying, bobbing, self-harm, and biting of enclosure bars. AI could allow us to offer alternative but similar forms of entertainment to zoo-goers. The Zoo of the Future is an interactive VR experience in which visitors can experience the natural world up close without enclosure bars or fences. The technology allows visitors to live the experience of walking right next to the animals and seeing birds fly over their heads. As this kind of technology becomes more advanced and more common and is marketed as an ethical alternative to zoos, we could see a reduction in zoo visits which would reduce the demand for keeping animals in captivity. The same kind of technology could be used for aquariums, with the potential, in cases of more advanced systems, to allow people to experience swimming with aquatic animals without ever having to keep them captive.

Already, animals are used much less frequently in films and TV shows than they once were, often replaced by CGI animals instead. However, animals are still used for filming – horses are especially difficult to replace due to the need for them to be ridden on screen. It is plausible that AI could enable us to replace all animals in TV and movies in the future, reducing stress for the animals involved.

Conclusion

In this article, I have explored what success could look like for animals if AI development goes well. The scope is deliberately broad, covering broader changes such as moral circle expansion as well as more specific ideas, from precision livestock farming to veterinary care and vivisection. My aim was to paint a picture of what the future could look like if the systems that cause so much animal suffering were changed, reshaped, or, in a best case scenario, made obsolete.

What so many of these AI applications have in common is that they allow us to see what was previously hidden. Animal suffering and death is made visible to investors, to consumers, to the public, and to regulators. Exposing suffering is certainly an important step, but that alone cannot guarantee change. None of the positive outcomes detailed here are certain; AI is a tool, and the outcomes of its use reflect the values of the user. Without substantial effort from researchers, animal welfare advocates, funders, and policymakers to make animal welfare a priority, the same AI developments that could vastly improve conditions for animals could do the opposite, intensifying factory farming, accelerating habitat loss, and potentially locking-in anti-animal values that we cannot amend. 

AI is already a key cause area in EA, as is animal welfare. It is worth investing additional time and energy into considering how these causes intersect, and understanding how important it is, for animals, that we consider their wellbeing now at this critical technological juncture. I expect that none of the outcomes included here preclude the others; many – though not all – of these outcomes could happen in parallel. There are almost certainly future AI applications that I have missed here, but the fundamental message is present nonetheless – we could shape AI to create a future for animals that is either a utopia or a hell. I hope that, by considering some of these positive outcomes, we will feel motivated to work towards them and develop strategies to achieve that goal. I would love for more researchers to focus on this intersection, and for funders to offer support to steer us towards a bright AI future for humans and nonhuman animals alike.

In Saving Animals, Saving Ourselves, Jeff Sebo writes that ‘in a perfect world, we would create a multispecies society in which all animals can flourish.’ We don't live in a perfect world but perhaps AI will offer us the tools to create one.

This article was written as part of the Electric Sheep Futurekind Fellowship. I used Claude Opus 4.5 as a writing assistant for brainstorming and feedback on drafts; all writing is my own.

[1] Based on a survey of 2000 UK adults

[2]  Copy of the relevant section of legislation: 

The Legislature finds and declares all of the following:

(a) Octopuses are highly intelligent, curious, problem-solving animals. They are conscious, sentient beings that exhibit cognitive and behavioral complexity, and are capable of experiencing pain, stress, and fear, as well as pleasure, equanimity, and social bonds.

(b) Octopuses have long-term memory and are capable of retaining information and recognizing individual people. The octopus carries out extensive foraging trips and uses landmarks to navigate the course.

(c) Octopuses have a well-developed nervous system, large brains relative to their body size, and a high level of problem-solving ability. They are known for their ability to learn, use tools, and exhibit behaviors that suggest a level of consciousness. Octopuses also display flexibility in their responses to different situations, which is indicative of cognitive complexity.

[3] https://link.springer.com/article/10.1186/s40813-019-0132-y; https://www.sciencedirect.com/science/article/abs/pii/S0167587716304688; https://www.cambridge.org/core/journals/animal/article/abs/reasons-and-risk-factors-for-beef-calf-and-youngstock-onfarm-mortality-in-extensive-cowcalf-herds/20D600263482DF1E78862E2E342DF6FC; https://www.researchgate.net/publication/324242805_On-farm_deaths_of_dairy_cows_are_associated_with_features_of_freestall_barns 

[4] I will use the term ‘pet owners’ throughout as it is most commonly used, but I personally prefer the term ‘caretaker’ or ‘carer’ as I do not consider pets to be property in a traditional sense.