From slaughterhouses to zoos to our own living rooms, AI-enabled interspecies communication could force us to finally confront what animals have been telling us all along.

Cross-posted from my Substack: https://substack.com/home/post/p-190377090 

It’s early morning and the zoo is not yet open. A keeper stands in front of the African elephant enclosure, watching as the matriarch, Asha, emerges from her indoor shelter. The sun lies low, glints of light peeking through the thick branches of the baobab tree. Cameras and microphones are scattered around the enclosure, hidden in corners and looking down from high posts. Asha moves towards the keeper in a straight line, making a low rumbling sound as she moves. There is a muffled murmur coming through the keeper’s earpiece. Perhaps he didn’t feed them enough the night before and she is expressing her displeasure. He uses an index finger to press the earpiece deeper into his ear and listens intently. He hears the words – unmistakeable this time: ‘I don’t want to be here.’

This might sound like science fiction, but AI-enabled interspecies communication may be closer than we think. Organisations such as Project CETI, Earth Species Project (ESP), and Interspecies Internet are working on facilitating communication of this kind through the collection and decoding of vast amounts of animal communication data. ESP states that their goal is to ‘decode and amplify the voice of nature and rebalance our relationship with the natural world.’ The creation of a future in which this kind of rebalancing would be possible would require both that we want to hear what animals have to say, and that we would be open to allowing those voices to shape our actions. In this article, I will discuss the significance of hearing animal voices in three distinct contexts: farmed animals, zoo animals, and companion animals. Each of these scenarios presents a unique challenge to our capacity to respect the voice of the other.

Currently, most interspecies communication research focuses on animals such as whales, dolphins, primates, or even dogs. In many cases, the choice of species to focus on aligns with what we know about the intelligence of various species. However, there is a notable absence of focus on the most exploited land animals on earth – pigs, cows, chickens. This is unlikely to be attributable to technical limitation; pigs are considered among the most intelligent animal species, demonstrating an ability to interpret a mirror image to find food, use tools, and solve puzzles. They are also highly social, communicating emotional states with many distinct sounds, and showing clear social preferences within their group. This gap likely reflects where funding and public interest flow rather than being attributable to researcher indifference, but that only shifts the responsibility. Our collective and selective curiosity reveals something uncomfortable about our capacity for respect for other species: we may only be willing to hear the voices of animals we want to listen to and acknowledge.

Interspecies communication technology will not necessarily offer us entirely new information about what animals want – in many cases we already know that. We have seen them struggle and attempt escape, and we have heard their screams. The difference is this: the technology could make their preferences - which we often hand-wave away as instinct - linguistically legible, removing the ambiguity that has, thus far, offered us a kind of moral protection. When animals communicate in a way that we cannot deny or ignore, we may be forced to reckon with uncomfortable truths. Of course, it is also possible that some animals would express contentment and that would be meaningful too, but a reluctance to ask would suggest we already suspect what we might hear. If we want to take seriously the idea of communication with other species, we are forced to ask ourselves: Are we ready to hear, and take seriously, a non-human ‘no’? This article argues that an important ethical shift of interspecies communication will come, not from what animals tell us, but from whether we are willing to treat their explicit refusals as morally significant.

The Technology

Although the idea of interspecies communication can sound like a Doctor Dolittle fantasy, numerous organisations have already begun decoding animal communication systems. Project CETI, a non-profit founded in 2020, states that their goal is to ‘[apply] technology to amplify the magic of our natural world’. They utilise advanced machine learning systems and robotics to record and translate sperm whale communication. In order to do this, they have created a large-scale dataset of sperm whale behaviour and vocalisations, which they use to train their AI technology to translate the communication of these whales. Crucially, the CETI team have uncovered a ‘sperm whale phonetic alphabet’ by linking behavioural patterns to vocalisations. The prospect of human-whale communication is not new. In 1967, almost sixty years ago, Dr Roger Payne captured recordings of humpback whale vocalisations and, after listening to them on repeat for days, noticed that whales don’t just call out but sing to each other in repeating patterns. He later referred to whale sounds as ‘Whalespeak’. More recently, in 2021, a team of scientists off the coast of Alaska played a humpback whale contact call through an underwater speaker and were shocked when a whale they had named Twain approached the boat. To keep Twain engaged, they attempted to match the latency of her calls to their own. If Twain waited 10 seconds, they did too. The interaction lasted for 20 minutes before Twain swam away. It is considered by many to be the first ever human-whale ‘conversation’ in whale language.

Earth Species Project (ESP), founded in 2017, also utilises state-of-the-art 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. There is also Interspecies Internet, a think-tank that works to accelerate interspecies communication; the project’s core 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.’ The technology is advancing faster than many realise. With so many species on earth, many of which are widely considered intelligent, the challenge of understanding animal communication is vast. Inevitably, certain species will be prioritised over others. How will decisions be made about which animals we should aim to understand first?

In wildlife conservation, it has long been acknowledged that funding predominantly supports the protection of charismatic megafauna, such as lions, tigers, and elephants. Charismatic species are generally large in size, aesthetically appealing, and mostly mammals. A 2025 analysis of 14,600 conservation projects over a 25-year period found substantial taxonomic biases in funding, with 29% going towards species of least concern, and just 6% of threatened species supported by conservation funds. Interspecies communication research faces different considerations than conservation; practical factors like communication style and cognitive complexity will inevitably shape which species are prioritised. But practical factors alone don’t explain the pattern. The bias in favour of charismatic animals cannot be dismissed as an influential factor. The following questions could be asked when assessing which animal species to focus our efforts on:

1. How intelligent are they?

2. Are their communication patterns conducive to mass data collection?

3. What impact would enabling understanding with this species have?

4. Which animals are we most interested in hearing from?

The first two address practical concerns. Breakthroughs in interspecies communication are more plausible, certainly in the near-term, in cases where animals already express themselves in ways that facilitate data collection, such as physical behaviours and vocalisations. Humpback whale song can be captured at scale with hydrophones; primate gestures can be recorded and catalogued. Already, many animals communicate in ways that we can make sense of – they cry out, protect their offspring, show affection, scream, charge, hide, freeze, beat their chests, stamp their feet, and all other manner of actions not so dissimilar from our own. This brings to light an important point, especially for those sceptical about the possibilities of this technology. Interspecies communication is unlikely to involve translating an animal’s thoughts or vocalisations into English sentences, as though an AI were reading the animal’s inner monologue. It seems unlikely that animals have an inner monologue structured in ways that are easily interpretable to humans. The exact form that the communication will take remains uncertain - outputs could be sentences, colours, scores, or something we have not yet imagined. In the next section, I will further explore what animals could tell us, and what they couldn’t.

For clarity, I will use sentences throughout this article to represent animal sentiments, but the ethical stakes remain the same regardless of the form of output: what matters is that the communication is explicit enough to remove ambiguity we currently rely on. This forms the basis for the reflections in this article. At the moment, we can dismiss even recognisable animal behaviour as instinctive, trivial, or unconnected to emotional states. We do it more easily with a lobster, of whom we have minimal understanding. It is more difficult with our dogs and cats. Reason indicates that their cries reflect unhappiness or pain, but still there remains some uncertainty that offers us moral protection. How would the world be reshaped if that were removed? How would it reshape human-animal relationships?

What Could Animals Actually Tell Us?

Regardless of what animals might be thinking, there are animal thoughts that we will likely never have access to. Understanding animals requires interpreting behaviour, movements, body language, vocalisations – if a thought is never outwardly expressed, even the most advanced AI systems would probably not have any way of accessing it. This applies to humans just as readily as it applies to non-human animals. Think about a human infant, for example; we have access to the thought ‘I am upset’ because, when upset, a baby is likely to cry, squirm, scrunch up their face, or other manifestations of this thought. If, however, an infant had the thought ‘I prefer this colour (green) to that colour (red),’ we would never know. A thought that is not outwardly expressed is not open to external interpretation of any kind. With that in mind, we can classify animal statements into three epistemic tiers.

The first tier comprises statements about acute affective states – an animal’s thought that they are excited, stressed, fearful, or similar. This tier contains the statements that are most accessible to us, and most likely to be subject to AI-enabled interpretation. An animal might scream, chirp, sway, growl, paw the ground, throw their head back, run, charge – these behavioural manifestations of the inner statement allow us access. Though these are the most accessible, they are also the least revolutionary. We already get a good sense if an animal is excited, afraid, or in pain – AI-enabled understanding here removes a degree of doubt but does not tell us anything completely unexpected.

The second tier is made up of continuous or consistent aversions and preferences; this would include statements such as ‘I do not want to live here,’ or ‘I am unhappy to be in a cage’. These statements may be accessible – though less so than those in tier one – and could potentially be more ethically disruptive. A statement such as ‘I do not want to be confined in this enclosure’ confronts us in a way that ‘I am unhappy’ does not, removing additional ambiguity.

Finally, the statements in the third tier relate to narrative or relational inner states. These are the least accessible, even with the assistance of advanced AI methods, as they are not behaviourally expressed – or, at least, not in a way that is interpretable by us as distinct from statements in tiers one or two. Statements such as ‘I miss my family,’ or ‘I want to return to the jungle’ belong to this tier. This is not to say that animals are incapable of thoughts of this kind; it is plausible that at least some species may have an inner life richer than we ordinarily assume, even if not structured like our own. Rather, the argument is that we may never know of these animal thoughts. They are hidden from us, forever uninterpretable through the opacity of the interspecies veil.

Three Reckonings

If we develop the technology to communicate with and understand animals more clearly, we will be forced to face uncomfortable questions across a wide variety of contexts. Here, I will examine three of them. Firstly, I will look at the implications for farmed animals, considering whether we will ever allow their voices to be heard and what it would mean, economically and morally, for us to take those voices seriously. Then, I will turn to animals kept in zoos. Animals are kept in zoos for profit, but many zoos also run conservation programs. If we knew, with little doubt, what the animals in our local zoo wanted, how much of our treatment of them would need to change? Finally, the spotlight will shift to some of the most beloved animals in our society – our companion animals. If our cats and dogs could tell us what they wanted - where they wanted to live and how they wanted to spend their time - the limits of our love for them may be brought to light.

Reckoning 1: Farmed Animals and Selective Silencing

It’s almost 3am but work in the slaughterhouse never slows down. It’s as noisy and chaotic as it is at any other time of day. Carcasses hang on the shackle line, ready for the scalding tanks to remove their hair. Another truck full of sows just pulled in, and the pigs are being ushered off the truck and into the slaughter line. Although much of the work is automated, operations are overseen by human supervisors. One of them spots a device around one pig’s neck. He recognises it – they use them to track welfare scores on certified high-welfare farms but they’re always taken off before they’re loaded onto the truck. As he moves towards the pig, he hears the device beep. He grabs a pair of shears to remove the collar and looks down as his fingers edge under the collar. Text lights up the screen: ‘I am afraid’.

A scenario like this is likely implausible. This is not necessarily due to technological limitations, but because it is improbable that we would ever allow the voice of a pig to be heard. In the scene described, a device – usually used to achieve a certain welfare tier – is left on accidentally. This is a crucial point. It is through human error, not intention, that we get an insight into what the pig is feeling. The suppression of certain voices to maintain the status quo has a long and violent history. Women’s voices present the perfect example. Ancient women were actively excluded from public speech. For much of history, women and girls did not have access to education. As recently as 1840, 60% of women in the UK were still illiterate. Without an ability to read or write, the opportunities for a woman to have her perspective widely heard were severely limited. Attempts to stop the suffragettes from speaking included violence, mockery, and institutionalisation. In 1914, suffragette Dorothy Evans went to speak at Queen’s University in Belfast. The male students did everything that they could to prevent her being heard including yelling, setting off a stink bomb and producing a mocking effigy of a suffragist. As Mary Beard says, ‘When it comes to silencing women, Western culture has had thousands of years of practice.’ This kind of silencing is intentional and political. Amplifying the voices of the marginalised functions as a confrontation and a call to action. My point is not that the experience of being silenced is the same in the historical case of women as would be for nonhuman animals in the future, but that both cases have a shared structural pattern, where those in power deliberately withhold the means of expression for those whose voices would threaten the status quo.

If we develop the technology to understand a pig or a cow, and we refuse to ever use that technology on the animals we exploit while freely using it on our companion animals, what would that say about us? Defenders of animal agriculture might argue that even increased clarity of communication would change nothing because food production already involves unavoidable harm. It seems that two distinct paths forward are possible.

Selective silencing: Let us imagine that we develop the technology to better understand our pets. Sensors, cameras, or microphones track our dog’s behaviour and affective state, and we are offered insights into how he is feeling at any given moment. It is not unrealistic to imagine that this kind of technology will develop. Already Petpuls Lab offer customers a collar that, according to their claims, monitors a dog’s activities, physical condition, and emotions. How readily would we apply this technology to pigs, for example, who are widely considered to be more intelligent than dogs? We have a lot more contact with livestock animals than with animals such as lions and tigers, so it is plausible that hearing from pigs could offer potentially impactful insights. If the technology is available and not cost-prohibitive, then it could be used, even intermittently or sparingly, on livestock animals. If we consciously choose not to ever use the technology to understand a pig in the animal agriculture industry, then this intentional silencing indicates that we are not ready to hear a perspective that could threaten the status quo.

Confronting what we hear: If we decide that, actually, it seems unethical not to utilise this useful technology to gain further insights into the lives of the animals that are entirely dependent on our care, then we will be forced to reconcile with what is likely to be an uncomfortable reality. Consider the scene described above. If we know, without doubt, that a pig is afraid, we have removed the excuse of ignorance and the moral protection offered by even this vaguest sense of uncertainty. The moral weight of that shift is significant, and the challenge deepens further if we consider that some livestock animals may be more cognitively complex than we assume. A recent case brings to light the possibility of a kind of cognitive complexity in livestock that we previously thought impossible. An Austrian cow named Veronika has been observed using tools – using both ends of a broom to scratch hard-to-reach places; according to scientists, this is considered multi-purpose tool use and it is exceptionally rare to see in non-human animals. As stated in the BBC Science Focus report, ‘these findings suggest that our assumptions about the intelligence of livestock and other animals may not be due to how clever the animals are, but rather whether they have been given the chance to explore their intelligence, and if humans have taken the time to notice’. Perhaps it is possible to imagine that livestock animals could tell us not only that they’re afraid, but that they don’t want to be killed. What would it mean to ignore someone’s plea not to be killed? There seems to be a difference between causing harm to someone who seems not to want it, and causing harm to someone who has stated, unequivocally, that they do not want it. The animal agriculture industry is, fundamentally, at odds with a respect for non-human refusal.

Reckoning 2: Zoos and Non-Human Consent

Let us return to Asha, the African elephant from the introduction, and consider how deeper communication with her could demand substantial changes in zoo operations – that is, if we’re willing to listen and get on board. Elephants have demonstrated advanced cognitive skills, including using tools, stopping calves from entering dangerous areas, recalling migration routes even years later, and digging wells when water is scarce. If a zoo cared about animal welfare then they may be required to take seriously the explicitly expressed preferences of the animals that reside there. Asha might be unhappy with her food and request something different. She may refuse veterinary care. She may not want visitors on a given day, preferring that she and her family be left alone. Taking the last example as a jumping-off point, let’s consider what a zoo that takes seriously the requests of its inhabitants might look like. In such a case, animal preferences, from food choices to willingness to accept visitors, have economic weight. Not all animals would be suitable candidates to participate in such a structure due to the limitations mentioned in the technology section, but it is certainly plausible that large mammals such as elephants, tigers, or primates, would. People would be disincentivised from visiting the zoo to some degree, due to the inconsistency regarding which animals can be seen on a given day. The zoo prices, on days where all animals are on display, would have to be much higher than they are now to allow prices to drop substantially on days when few large animals are willing to accept visitors. An app might show potential visitors the available animals each morning, with the amended price listed so users can make an informed decision. However, even this does not fully capture the strangeness that full respect for animal perspectives could have. If we believe that animal consent has some value, albeit weaker than that of an independent adult and closer to that of a child, then respecting that consent means that it can be withdrawn at any time. A button or lever could be installed to allow an animal to close the viewing area, but, on an economic level, such a system is not sustainable. Animals could be offered incentives, such as treats, in exchange for willingness to be seen, but this would raise questions about coercion. It is also possible that animals would tell us, again and again, that they are not happy in their enclosure. Would we then have a moral responsibility to move them to another facility or set them free?

Zoos are more likely than the animal agriculture industry to engage with this technology. It could offer them the chance to improve the public perception of their animal welfare standards, forge research partnerships, and offer visitors new and exciting opportunities to engage with the animals in their own language. With that in mind, it’s reasonable to think that, sometime in the future, interspecies communication and understanding with zoo animals will happen. What we decide to do with what we learn becomes the real question. As discussed already, animals already communicate with us, and this includes zoo animals. Animals in zoos can display psychological distress in captivity, including behaviours such as swaying, pacing, chewing the bars of their enclosure, or rocking. Animals have attempted escape. They have refused food. In reality, a lever that shuts off their enclosure from the viewing public is feasible even now, without enhanced communication, but it is not something we offer. We can justify this to ourselves by citing uncertainty about an animal’s desires or emotional state, or perhaps uncertainty around whether or not they even have the cognitive capacity to understand such an arrangement. What interspecies communication changes is the strength of these justifications. When we remove much doubt about an animal’s unwillingness to remain in captivity, we could be forced to face either letting them leave or keeping them prisoner against their will – something we may be doing now, but with plausible deniability.

Zoo-based conservation projects potentially create another difficult conflict. Most zoos mention conservation efforts in their mission statement. The function of zoos when it comes to wildlife conservation involves practice, advocacy and research. Within these, zoos practice methods such as captive breeding and species reintroduction programs, as well as advocacy and education efforts involving public engagement, promoting stewardship, and fundraising. Finally, conservation research may be focused on wildlife biology, population dynamics, animal behaviour, and health. The question arises regarding what we do when animals tell us, with certainty, that they do not want to participate in projects that are for the good of their species long-term. Can we carry out research on animals that are unhappy participating, knowing that it is in service of a greater goal?

Here, we might look at how consent to medical intervention is handled for minors. In the UK, Gillick competence is used as a standard to help determine whether a child is capable of consenting to or refusing medical care. In assessment, a professional assesses factors such as: the child’s mental capacity, their understanding of the issue and what is involved, their understanding of the risks and consequences of their decision, and their ability to rationalise about their decision. Medical intervention requires a high bar, as the stakes can be life or death, so consent or refusal to engage in conservation research may demand a lower level of cognitive reasoning. Nonetheless, it is plausible that ethically sound research could be carried out on animals, even if they are unhappy about their involvement, due to their potentially limited ability to comprehend the stakes of their decision. Research that causes minor or temporary harm in the short-term may be defensible if long-term gains are substantial for the relevant species. However, we should treat this conclusion with due suspicion, as humans have a long history of exploiting marginalised groups ‘for their own good’ in ways that are later considered blatantly unethical.

Reckoning 3: Companion Animals and Guardianship

You have just returned from a holiday abroad. After hauling your suitcase off the carousel at the baggage claim and loading it into the boot of your car, you drive towards your mother’s house to collect your terrier Alfie. Your mother loves Alfie and is always thrilled to take him when you go away. You pull into the driveway, hop out, and ring the doorbell. Your mother opens the door and you see Alfie standing in the hallway behind her. He’s happy to see you, but begins to back away when he spots the crate in your hand. Your phone beeps – a notification from your pet communication app. You open it and see the following message next to the photo of Alfie’s cheeky face: ‘I want to stay here.’

The situation with companion animals is characterised by both the lowest material stakes and, simultaneously, the highest emotional stakes. Companion animals are less exploited for economic gains than animals in farming or in entertainment, although there are financial incentives related to the breeding of companion animals. It is plausible that a cat or a dog could express an unwillingness to be held, rubbed, or go for a walk, an unhappiness about their living situation, or a refusal to wear a coat or get into the car. In the current context, we override many of the desires expressed by our companion animals, telling ourselves that we are acting in their best interests. Enhanced interspecies understanding could make this override more obvious, and call into question what it means to act in their best interests while also ignoring their expressed preferences. Our companion animals may be delighted to live with us, and could be entirely content to live in exactly the way that we hope they will. This is certainly possible – many people treat their companion animals as family members, caring for them and doting upon them on a daily basis. However, we cannot take this for granted.

It may be the case that our companion animal does not want to live with us. Our dogs might express a desire to stay with the neighbour or family member who took care of them while we were away. Our cat may want to leave the house and never return, which raises the question of whether we should go looking for someone who does not want to be found. Could a cat survive without us? In such a case, are we retrieving them from the cold outdoor world for their benefit or for ours? We are guardians of our companion animals, and enhanced interspecies understanding could force us to confront what ethical guardianship looks like. When we genuinely love our companion animals and want to protect them, we need to consider how much weight to give to their expressed desires compared to our knowledge about what is optimal for their welfare. With a cat who wants to roam freely outside, we can certainly make the case that refusing that want is ethically justifiable. There are substantial risks for a cat living outdoors without a guardian, including exposure to harsh weather, predators, and disease. We have domesticated animals such as dogs and cats, such that it can be difficult for them to survive without us. However, the case is less clear if our dog or cat wants to stay with someone else, as in the example of Alfie detailed above. If, for argument’s sake, your mother was willing to keep Alfie, how could you justify bringing him home? He is surely not at any increased risk living there compared to living with you. Possibly, the only justification you could make for overriding Alfie’s wishes in this case is that you want to have him home with you. To bring him home would mean putting your own wishes over his, forcing you to confront the limits of the love you claim to have for him.

Preparing for the Reckoning

During World War I, Hugh John Lofting was haunted by his experiences witnessing terrible suffering on the battlefield. He particularly noticed that the horses – used for carrying troops, heavy artillery, ammunition, and medical supplies – were abused and overworked in horrific conditions. He wrote to his children back home, not about the atrocities he witnessed, but about a doctor who treated animals, and learned their languages to best care for them. In The Story of Doctor Dolittle, using Polynesia the parrot as a mouthpiece, he wrote ‘People make me sick. They think they’re so wonderful. The world has been going on now for thousands of years, hasn’t it? And the only thing in animal language that people have learned to understand is that when a dog wags his tail he means ‘I’m glad’!’ Lofting’s fantasy was one of enhanced understanding, of an ability to fully empathise with non-human animals and to make sense of their suffering. Although, as stated already, AI-enabled interspecies communication is unlikely to lead to a Doctor Dolittle equivalent fantasy world, it is likely to offer us opportunities to increase our level of understanding and extend our compassion accordingly.

If we are unprepared to hear a non-human ‘no’, then it raises the question of whether or not we should work on developing this technology at all. From my perspective, even with the associated complexities that are sure to arrive, it is worth pursuing. If we could understand animals, really understand them without the notable ambiguity that now exists, it would force a shift in our thinking. This shift may be gradual rather than instantaneous, but eventually the sound of non-human pleas for change, for consideration, and for justice would be too loud to ignore and we will emerge transformed. This is the position of hope. History, however, suggests that we are infinitely capable of ignoring what we do not wish to hear, even when the evidence is undeniable. The technology removes the excuse of doubt, but it does not remove our capacity for wilful ignorance. We cannot, once ambiguity is removed, dress intentional ignorance up as uncertainty, and perhaps that is sufficient as a starting point.

Few changes could offer a more significant push for moral development and growth than our ability to hear the voices of the voiceless. If we decide, based on our unwillingness to hear non-human perspectives, not to pursue the development of this technology, that in itself acknowledges a moral deficiency in us that demands further reflection. Most people consider themselves to be morally superior to others, with a view of themselves as ‘just, virtuous, and moral’. If we push back against technological development precisely because it would force us to acknowledge the perspectives of the exploited, then our assurance of our own moral goodness deserves closer scrutiny. We should embrace our chance to live up to our vision of ourselves as good people by opening our minds. We could strive not merely to understand Asha, Alfie, and the unnamed pigs in the slaughterhouse, but to make decisions with their preferences, desires, and fears considered and accounted for – it is not enough to make perspectives legible, we also need to care about what we hear.

As Aza Raskin once said in an interview with Adam Grant: ‘We don’t change when we speak. We change when we listen.’

Disclosure: I used Claude Opus 4.5 as a writing assistant for brainstorming and feedback on drafts; all writing is my own.