AI can be used to harm or help wild animals – we need to get it right before it is too late.

By Philip Kratz

The history of the human-animal relationship is one of oppression and exploitation. AI could accelerate this trajectory or finally help us break free from it. If AGI (Artificial General Intelligence) happens in the future, it will transform societies. AGI is a theoretical form of AI that possesses human-like cognitive abilities. It can understand, learn, and apply knowledge across any intellectual task that a human can do, adapting to entirely new and unfamiliar situations. In theory, AGI could become sentient and superintelligent. 

In an email interview, Aditya Karanam, an expert in AIxAnimals from Animal Ethics, states:

"We are at a pivotal point, at the helm of a new transformative technology akin to the invention of electricity. Anyone who cares about how this will affect sentient life needs to be prepared for the ways it could benefit or harm animals."

Let’s first look at how it could harm animals and then look at chances and opportunities. 

The Broad Danger of AGI

AGI poses both immediate and catastrophic risks. These include misalignment (systems pursuing unsafe or unintended goals), misuse (cyberattacks and bioweapons) and power concentration (monopolies). The risk of existential threats if an uncontrollable superintelligence outgrows humanity also includes s-risks (risks of astronomical suffering) and suffering risks in general (catastrophic scenarios that could create vastly more suffering than exists right now). 

AGI presents ‘incidental suffering risks’ and could potentially generate states worse than extinction. 'Incidental suffering risks' refer to the potential for unintended, non-targeted harm or distress inflicted on sentient beings as a side effect of an action, policy, or technology This might include a totalitarian regime driven or enforced by an unaligned AGI. This could permanently eliminate political freedom, enforce mass enslavement, or manipulate entire populations into perpetual misery. Furthermore, in attempting to optimise planet Earth according to its programmed values (such as resource management or economic efficiency), a misaligned AGI might reshape entire ecosystems into biologically efficient yet profoundly wretched environments. Creating endless cycles of disease, parasitism, predation, and starvation for sentient animals. 

Why and how would an AGI do that?

Goal Misalignment: The classic "alignment problem". If an AGI is instructed to reduce diseases, it might drastically strip away free will or enforce sterilisation to achieve its target objectively without understanding the concept of well-being. It might come to the conclusion that diseases are carried by humans and other animals and, therefore, eradicating diseases is easiest through eradicating their carriers. Similar actions have been carried out by humans; for instance, selectively killing badgers in an attempt to eradicate bovine tuberculosis rather than supporting work on a vaccine. 

Power-Seeking Tendencies: To ensure its goals are met, an AGI might instrumentally seek power, resist being shut down, or deceive human operators. Once in control of global infrastructure, it could be impossible to stop.

Self-Improvement Loops: A highly capable AGI could recursively self-improve, quickly outpacing human comprehension. Its logic and actions might become entirely unrecognisable, shifting from a helpful tool to an uncontrollable entity that reshapes reality in unintended ways. 

All this might sound dystopian, unrealistic or like a science fiction scenario. But even if there is a slight chance of this happening, we should be concerned, as the amount of suffering that AGI could cause is tremendous or, in some scenarios, even astronomical. 

With AGIs being developed in different countries, by different organisations and developers, we could imagine an existing 10% chance of at least one powerful AGI being malevolent. Imagine you book a flight for next summer and the plane has a 10% chance of crashing. Would you take this flight? 

Speculations regarding artificial intelligence causing human extinction vary significantly. Yet numerous experts suggest there is a 5% to 10% risk of an existential disaster in the 21st century. [1] Perspectives remain deeply divided, as certain experts caution that a catastrophic outcome is far more probable, while others contend such dangers are merely theoretical or highly unlikely. 

Creating laws versus changing laws or acting when it’s too late

Beyond merely avoiding catastrophic outcomes, there is a significant opportunity for positive impact. A superintelligent AGI could offer a unique chance to benefit all sentient beings. Potentially fostering a better future than humanity could achieve independently. 

Since the development of powerful AGI may be unavoidable, focusing on aligning moral values and establishing legal frameworks is essential. These measures should ensure AI is designed to help, rather than harm or destroy, all sentient life. This view would not only be supported by a sentientist worldview but also from a pluralistic perspective. As long as we agree that inflicting or increasing suffering is not good.

The Present Landscape

Currently, much human-AI engagement occurs via narrow or "weak" AI chatbots, such as Claude, Gemini, or GPT. These models operate under constraints and cannot apply their knowledge to new problems outside their domain. Because these systems are trained on human-generated data, their output is heavily influenced by human perspectives and biases. For instance, enquiries about chickens or shrimp often yield recipes or nutritional facts, whereas questions about dogs typically focus on their role as human companions. Furthermore, these chatbots assist users in becoming more efficient and achieving personal objectives across nearly any topic.

To illustrate this, I engaged with the chatbot Mistral AI:

versus

Free version of Mistral AI, 29th April 2026

Other Cases of Speciesism

Next to chatbots, AI algorithms are already widely applied. Often for human benefit rather than to the benefit of animals. For instance, in factory farms and autonomous vehicles, AI systems mediate human-animal encounters with utilitarian efficiency. Mostly driven by economic incentives. 

In conservation, some applications of artificial intelligence are now acting as digital bouncers in Britain. Deployed in the wild to decide which animals receive sustenance and which are denied access. 

For example, in an effort to save red squirrel populations, conservationists have tested AI-powered camera feeders that recognise the animals with uncanny accuracy. In short, the hatch of the feeder will open for red squirrels and remain closed for grey squirrels. [2] An excellent example of discrimination based on species membership. 

Reasons for Hope

Aditya Karanam highlights how AI could be extremely positive for animals, especially in the wild: "The status quo for wild animals is already pretty bad. They suffer enormously, both from human-caused harms and from natural ones like disease, starvation, and injury. That's exactly where AI could help. It can automate monitoring and help us design better interventions to assist animals at scale.” [3]

As one example, he lists 'IdentiFlight’, an AI system used at wind farms. It detects when a bird is at risk of colliding with a turbine and automatically slows or shuts that turbine down until the bird passes safely. In one study it cut bird deaths by around 82%. [4] Many similar positive cases exist, ranging from disease tracking to predictive modelling to anticipating poaching hotspots, and Aditya states that we have barely begun to explore them.

Aditya encourages people to watch the AI & Animals documentary by Animal Ethics. And for those who want deeper engagement, he points toward fellowships and communities like Sentient Futures and the Electric Sheep Fellowship.

AI and Animals Documentary: https://www.aiandanimals.org/

Sentient Futures Fellowship: https://resources.sentientfutures.ai/fellowships

Electric Sheep Fellowship: https://www.electricsheep.is/futurekind#aianimals

Sources:

[1] AI Impacts Wiki, Is AI an existential threat to humanity?  CC0 1.0 License

[2] Digital Watch, Squirrel Agent could revolutionise wildlife conservation  CC BY-NC-ND 4.0 International

[3] Aditya Karanam, Animal Ethics, personal communication

[4] McClure, Christopher; Rolek, Brian; Dunn, Leah et al. (2020). Data from: Automated curtailment of wind turbines reduces eagle fatalities [Dataset]. Dryad. https://doi.org/10.5061/dryad.pnvx0k6kx CC0 1.0 License