Animal welfare advocates and AI researchers met in San Francisco in early February to explore whether machine learning could do what decades of traditional campaigning has not: protect non-human animals at scale.
The event took place at Mox, a shoes-free coworking space decorated with Persian rugs and mosaic lamps — an aesthetic the organisers appeared to choose deliberately to lower the barrier between cause-driven advocates and technically minded engineers. Wildlife conservationists, farm animal protection campaigners, and AI researchers shared the same room, a combination that would have been unusual even five years ago.
The Scale of the Problem Advocates Are Trying to Solve
The numbers framing the animal welfare movement are stark. An estimated 80 billion land animals are raised and slaughtered for food globally each year, according to the Food and Agriculture Organization of the United Nations. Wildlife populations have declined by an average of 69 percent since 1970, per the World Wildlife Fund's 2022 Living Planet Report. Traditional advocacy — litigation, legislation, public campaigns — has made limited headway against figures of that magnitude.
Advocates at the February gathering argued that AI could provide monitoring, measurement, and intervention tools operating at a scale that human researchers cannot reach. One wildlife advocate described computer vision systems capable of tracking animal behaviour across large habitats, work that would previously have required hundreds of field researchers operating over years.
The pitch was that animal welfare presents hard, unsolved problems that could engage serious researchers — and that the field's relative neglect means individual contributions could have outsized impact.
The event was organised in part by groups aligned with the effective altruism movement, which has identified animal welfare as a core cause area and has channelled significant philanthropic funding toward both AI safety research and animal advocacy.
Three Concrete Applications Circulating Among Attendees
Several specific use cases emerged from the gathering, according to reporting by MIT Technology Review.
Acoustic monitoring represents a near-term application. Machine learning models can already identify individual animals by vocalisation, track migration patterns, and detect signs of distress in livestock facilities — data that could support both conservation work and legal challenges to industrial farming.
Predictive modelling offers a second avenue. AI systems trained on historical data about disease outbreaks, climate shifts, and human land use could help conservation organisations deploy resources more efficiently, concentrating interventions where projected impact on animal populations is highest.
Video analysis of agricultural facilities represents the application with perhaps the most immediate policy relevance. Several countries, including the United Kingdom, have begun piloting camera systems in slaughterhouses. AI analysis could make that footage actionable for enforcement rather than merely archival.
Recruiting Engineers Who Could Work Anywhere
The harder challenge is cultural. AI researchers face no shortage of well-funded opportunities in commercial technology, defence contracting, and healthcare. Animal welfare does not carry equivalent professional prestige or financial reward.
Organisers acknowledged this directly. Speakers at the Mox event addressed the neuroscience of pain perception in fish, the social complexity of pig behaviour, and the methodological challenges of measuring subjective experience in non-verbal animals. The strategy — presenting animal welfare as a domain of genuinely hard, unsolved scientific problems — mirrors the early recruitment approach used by AI safety organisations, which similarly had to convince technically minded researchers that consequential but neglected problems deserved serious attention.
Several attendees had backgrounds in AI safety before moving toward animal-focused work, suggesting the talent pipeline may already exist in nascent form.
Friction Inside the Coalition
The alliance is not frictionless. Mainstream animal welfare organisations — those focused on anti-cruelty legislation or companion animal protection — largely operate outside the effective altruism orbit and have sometimes been sceptical of its utilitarian framing of animal suffering.
There are also disagreements about which animals should receive priority. Effective altruism-aligned groups have concentrated on farmed animals, given the scale of numbers involved. Conservation advocates emphasise wild populations and ecosystems. The tools best suited to one context may be poorly adapted to the other.
Data access presents a further structural problem. Detailed behavioural and physiological data on farmed animals is largely controlled by agricultural corporations with little incentive to share it. Building usable training datasets may require either regulatory intervention or independent collection efforts — neither straightforward to achieve.
No funding commitments or formal research partnerships were announced from the February gathering. Organisers described it as a relationship-building exercise, with similar convenings planned and several attendees expressing interest in longer-term collaboration.
What This Means
As AI systems become embedded in agriculture, conservation management, and regulatory compliance, whether animal welfare is built into those systems from the outset — or retrofitted later — will carry real practical consequences for billions of animals and the organisations trying to protect them.