CaML researches how synthetic pretraining data can shift AI systems towards greater compassion and moral open-mindedness regarding all sentient beings, including animals and potential digital minds.
CaML researches how synthetic pretraining data can shift AI systems towards greater compassion and moral open-mindedness regarding all sentient beings, including animals and potential digital minds.
People
Updated 05/18/26Co-Founder
Co-Founder
Funding Details
Updated 05/18/26- Annual Budget
- $159,000
- Current Runway
- -
- Funding Goal
- $159,000
- Funding Raised to Date
- $128,000
Org Details
Updated 05/18/26Compassion in Machine Learning (CaML) is an AI safety research organization dedicated to shaping how advanced AI treats all sentient beings. Founded in August 2024, CaML conducts research on aligning artificial intelligence with the well-being of all sentient beings before the technology outpaces the ability to guide it. CaML's core research method is Synthetic Document Finetuning, which generates targeted synthetic pretraining data to shift future AIs towards considering the welfare of non-humans, particularly animals and digital minds, in a way that is robust and designed to scale to superintelligence. The organization has demonstrated that this technique produces large improvements in animal compassion that persist even after supervised fine-tuning, with frontier labs expressing interest in such techniques to complement their existing alignment tuning. CaML also trains AIs to exhibit moral humility to avoid incorrigibility and value lock-in, encouraging appropriate uncertainty and risk-avoidance about what entities matter and how much. Their benchmarking work includes the Animal Harm Bench (AHB), MORU (Moral Reasoning Under Uncertainty, a 201-question benchmark across four datasets covering alien organisms, digital sentience, AI values, and human compassion), and the CompassionBench leaderboard hosted at compassionbench.com. The organization was co-founded by Miles Tidmarsh, a research economist with a Masters in Economics from the University of Melbourne and extensive experience in the Effective Altruism community, and Jasmine Brazilek, who brings over six years of cybersecurity experience including security systems design at Anthropic. The team also includes technical staff, a web designer, and several volunteers, with an advisory board that includes NYU philosopher Jeff Sebo, Sentient Futures founder Constance Li, and Open Paws founder Sam Tucker. CaML operates as a fiscally sponsored project of Players Philanthropy Fund, Inc. and partners with organizations including Sentient Futures, Electric Sheep, Open Paws, and Wild Animal Initiative. Their models and datasets are publicly available on HuggingFace, where they have published approximately 200 models and 92 datasets.
Theory of Change
Updated 05/18/26CaML believes that the values embedded in AI systems during pretraining are critical and persistent, surviving through subsequent fine-tuning stages. By generating high-quality synthetic pretraining data that encodes compassion towards all sentient beings, they aim to shift the baseline moral orientation of future AI models before capabilities outrun alignment efforts. Their benchmarks (Animal Harm Bench, MORU, CompassionBench) create measurable standards that frontier labs can adopt to evaluate and improve their models' treatment of sentient welfare considerations. By targeting the pretraining stage rather than post-hoc alignment, and by encouraging moral humility and open-mindedness rather than rigid value lock-in, CaML aims to produce AI systems that are robustly compassionate in ways that scale to superintelligence.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26
Research project and 2026 preprint by CaML that uses synthetic midtraining documents to improve language models’ compassionate reasoning about animals and introduces the ANIMA benchmark for evaluating animal-welfare reasoning.
A public leaderboard developed by Compassion in Machine Learning that evaluates how frontier AI models reason about animal welfare, built on the UK AI Safety Institute’s Inspect framework and extending earlier Animal Harm Benchmark work.
A benchmark released by Declan McKenna in collaboration with CaML that tests whether large language models’ compassionate moral consideration generalises across aliens, digital minds, and vulnerable humans under moral uncertainty.
Discussion
No comments yet. Be the first to share your thoughts.