By Alistair Stewart
How can we steer AI development today and in the future to reduce animal suffering, rather than increase it? One place to look is AI governance: the set of norms, policies, laws and institutions whose purpose is to influence how AI is developed and used.
Last year, Ronen Bar and I conducted an analysis of intergovernmental AI governance frameworks to investigate how current approaches to AI governance view animal interests. We found a systemic exclusion of animals’ interests. In the rare cases where animals are mentioned, they are usually seen as little more than a resource for human use, or as a critical component of the ecosystem, rather than an end in themselves.
This “animal gap” matters because AI may have a profound impact on how animals’ lives go in the future, and AI governance tools may help us shape AI so as to protect animals.
How things could go badly for animals
Inspired by Paul Christiano’s 2019 piece What failure looks like, Niki Dupuis and I sketched a range of ways in which a future with powerful AI could go badly for animals. In particular, we are worried about AI systems increasing the intensity, duration, irreversibility or scale of their suffering.
Below are three overlapping failure modes that particularly concern us, because they each risk “locking in” values, systems or outcomes that are harmful to animals. Each of these failure modes involves shifting the distribution of power, away from humans who want to protect animals, and towards humans or AI systems that are indifferent to their interests.
1. AI takeover with animal-indifferent values
A powerful AI gains long-term control over society’s key institutions. Reflecting widespread attitudes among humans, this AI has speciesist values that render it mostly indifferent to animal suffering. Even if humans continue to exist, the AI decides to prioritise the aesthetic value of nature over helping individual wild animals, many of whom endure suffering such as predation, disease and starvation. These values shape the world for as long as the AI remains in control.
2. AI-enabled human takeover with animal-indifferent values
A small number of humans use AI to orchestrate a global coup, effectively establishing a permanent global government. Even though some of these humans are concerned about the wellbeing of animals, most of them do not prioritise animals enough to protect or promote their interests in this new world order. A small number of animals are treated relatively well – for instance, as companions for the wealthy. The rest are severely mistreated, including through intensified AI-assisted factory farming. There is no scope for political opposition to this AI-backed global government, thanks to its advanced surveillance and law enforcement.
3. Disempowerment of animal advocates
Powerful AI causes mass job loss and economically disempowers all but a tiny elite of humans, few of whom are pro-animal. Animal advocates lose the bandwidth, energy, money and time to do anything except provide for themselves and their families; scarcity drives the general public to care less about animals. Alternatively, powerful AI could give us access to such stimulating, pleasurable and addictive sources of entertainment that we simply stop caring about animals, who continue to be harmed.
Meanwhile, industries that use animals for food, experimentation and other purposes use AI to create sophisticated and persuasive advertising campaigns that give them an unassailable advantage over animal advocates. The prevalence of deepfakes may discredit real footage of animal abuse, and erode the shared reality necessary for public debate.
What we might do
One way of protecting vulnerable individuals at the moment is to represent their interests and enshrine their rights in formal and informal governance mechanisms. For instance, today we have:
- Social norms that stigmatise child sex abuse
- Government policies for ensuring that disabled people have access to education
- Laws that protect some animals from mutilations (in some contexts)
- Institutions like governments and courts that uphold voting rights
These mechanisms provide an avenue to include animals’ interests in AI governance. More specifically, this might look like:
- Promoting social norms of caring for animals at companies developing AI
- Enacting policies that support the AI-assisted development of slaughter-free alternative proteins
- Using laws to ban certain forms of AI-assisted factory farming
- Intergovernmental institutions that incorporate explicit concern for animals in AI governance frameworks
At the highest legal levels, a foundation for protecting animals in AI governance already exists. Article 13 of the Treaty on the Functioning of the European Union requires that full regard is paid to “the welfare requirements of animals” as “sentient beings” in policies relating to technological development. The EU AI Act could be amended to reflect this: adding the protection of animals as a core value, including risks to animal wellbeing in mandatory assessments, and banning AI systems that directly inflict suffering on animals. At the United Nations, the High-Level Advisory Body on AI should also include dedicated representatives for animal interests.
As far as I can tell, Serbia’s 2023 ethical guidelines for AI are unique among nations in referring to concern for the wellbeing of animals. These guidelines recognise that animals can suffer, and state that AI systems must be developed “in harmony” with the wellbeing of animals as well as that of humans. This reference is vague and minimal, but we hope it can nonetheless serve as an example to other countries.
To conclude
AI governance tools can play a role in protecting animals in a future transformed by powerful AI, even if they are not sufficient on its own. Humanity will presumably also have to work out how to control the goals of AI systems, and what we collectively want those goals to be. Even so, making the norms, policies, laws and institutions surrounding AI more effective at reducing animal suffering is a step towards protecting animals and all sentient beings from suffering – another way of building our fortress.
If you have interests and/or experience in working on AI governance from an animal-friendly or suffering-focused perspective, please get in touch with the Center for Reducing Suffering – we would love to hear your ideas.