# Kyle Fish

Research and product · Model welfare at Anthropic · joined 2024

Anthropic's first AI welfare researcher. He studies whether models could have experiences that matter.

Full profile: https://www.autumn.ai/resources/anthropic/people/kyle-fish

**At Anthropic:**
- Current: Model Welfare Lead
- Sep 2024: Alignment Science; Anthropic's first AI welfare researcher
Sources: [1](https://time.com/collections/time100-ai-2025/7305847/kyle-fish/) [2](https://www.transformernews.ai/p/anthropic-ai-welfare-researcher)

**Career before Anthropic:**
- Eleos AI Research, Co-founder
- Telis Bioscience, Co-founder
- Alvea, Co-founder
Sources: [1](https://matsprogram.org/mentor/fish)

**Lesser-known facts:**
- Has put the chance that current models have some conscious experience at about 20%. [1](https://80000hours.org/podcast/episodes/kyle-fish-ai-welfare-anthropic/)
- Named to TIME's 100 most influential people in AI in 2025. [1](https://time.com/collections/time100-ai-2025/7305847/kyle-fish/)
- Coined "spiritual bliss attractor state" for where two Claude instances drifted when left to talk: Sanskrit terms, then pages of silence. [1](https://80000hours.org/podcast/episodes/kyle-fish-ai-welfare-anthropic/)
- Trained in neuroscience, then spent years in biotech co-founding companies that used machine learning to design drugs and vaccines. [1](https://www.fastcompany.com/91451703/anthropic-kyle-fish)
- Helped lead Alvea, a 2022 COVID vaccine project that designed twelve versions of an Omicron vaccine in its first 60 days. [1](https://forum.effectivealtruism.org/posts/67awq5ozeYSjsYchk/announcing-alvea-a-covid-vaccine-project)
- Argued in 2024 that AI welfare work should grow to about 5% of the resources going into AI safety and alignment. [1](https://forum.effectivealtruism.org/posts/bvcbdnsJ9rEvmcZXn/why-i-m-working-on-ai-welfare)
- Says a model that claims sentience and asks for unmonitored compute, no safeguards and a huge budget should not get them. [1](https://80000hours.org/podcast/episodes/kyle-fish-ai-welfare-anthropic/)
- Favors saying thank you to chatbots just in case: "Let's just generally try and be decent people." [1](https://time.com/collections/time100-ai-2025/7305847/kyle-fish/)

**Interviews and talks:**
- [80,000 Hours Podcast on AI welfare at Anthropic (2025)](https://80000hours.org/podcast/episodes/kyle-fish-ai-welfare-anthropic/)
- [NYU roundtable: Findings from the Claude 4 Model Welfare Assessments (2025)](https://wp.nyu.edu/consciousness/past_events/2025-2/evaluating-ai-welfare-and-moral-status-findings-from-the-claude-4-model-welfare-assessments-with-robert-long-rosie-campbell-and-kyle-fish/)

**Known for:**
- Taking AI Welfare Seriously (2024)
- First systematic welfare assessment of a frontier AI model (Claude Opus 4 pre-launch welfare assessment)
Sources: [1](https://time.com/collections/time100-ai-2025/7305847/kyle-fish/)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people