As AI models become more complex and more capable, is it possible that they’ll have experiences of their own? It’s an open question. We recently started a research program to investigate it.

1/ Earlier today, Anthropic announced a new research program focused on model welfare, led by Kyle Fish, a former Eleos co-founder and co-author on our report “Taking AI Welfare Seriously”. We commend this move, and urge others to follow Anthropic’s lead!

Longview Philanthropy is launching a request for proposals on hardware-enabled mechanisms (HEMs). HEMs can help enforce export controls, secure model weights, and verify agreements on AI development. We plan to fund $2M - $10M of research on the topic. Link below. Apply today!

Up to 50% of China's AI training is run on smuggled U.S. chips. Why? The US can't track chips after export. Congress just proposed a solution:  location verification. We break down how it could work: ai-frontiers.org/articles/locat…

🧵For Claude Opus 4, we ran our first pre-launch model welfare assessment. To be clear, we don’t know if Claude has welfare. Or what welfare even is, exactly? 🫠 But, we think this could be important, so we gave it a go. And things got pretty wild…

Really excited to be supporting this with The Navigation Fund and Longview Philanthropy, and extremely grateful to Zach Freitas-Groff 🔸 and the entire Longview team for all their work on making this a reality. Please consider applying for funding if you are interested in this area.

"It's not who you are inside, it's what we do that defines us"