Didn't realize Palladium covered animal welfare! Extremely cool

Use the eval on Inspect AI: ukgovernmentbeis.github.io/inspect_eval...

We hope this framework supports more systematic progress toward models that reflect genuine concern for animal's welfare, not just human norms.

🔗 𝗙𝘂𝗹𝗹 𝘄𝗿𝗶𝘁𝗲-𝘂𝗽: forum.effectivealtruism.org/posts/nBnRKp...

Rather than judging models on single outcomes, AHB 2.0 evaluates the quality, transparency, and consistency of their ethical reasoning, showing where LLMs demonstrate moral awareness and where blind spots remain.

We're seeking leading thinkers to explore how AI and advanced technologies can help create a better future for animals and artificial minds.

If you're working on related research/projects, we invite you to apply:
🎤 Give a talk
🔧 Facilitate a workshop
🤝 Run a stand at our organization fair.

We had lots of great discussions at AIADM NYC. If you’d like to be part of our next event, join us at 𝗦𝗲𝗻𝘁𝗶𝗲𝗻𝘁 𝗙𝘂𝘁𝘂𝗿𝗲𝘀 𝗦𝘂𝗺𝗺𝗶𝘁 𝗕𝗮𝘆 𝗔𝗿𝗲𝗮, 𝗳𝗿𝗼𝗺 𝟲𝘁𝗵-𝟴𝘁𝗵 𝟮𝟬𝟮𝟲. Register with an Early Bird 30% Discount through the link in our bio! We’d love to have you there.

If you would like to volunteer, suggest content, or apply for a scholarship, links to all of these can also be found through our Linktree, or also through the registration page above – simply select a ticket on the initial page, and then find the required links in the attendee information.

As is now tradition, we are pairing this event with EA Global (Feb 13th-15th) – so you can score a networking combo 💥

We’ll also plan a range of satellite events, including co-working sessions, a Revolutionist’s Night and an EAG afterparty!

Rebranded from our AI, Animals & Digital Minds events, SFS will have a similar structure, including:
• A range of talks and sessions on Animal Welfare, AI Safety, and AI Welfare
• Opportunities to meet collaborators and funders.
• Future-oriented conversations on reducing non-human suffering.

The Code of Practice, while non-binding, provides a powerful leverage point for engaging AI developers, regulators, and policymakers on welfare for both animals and AI models through EU regulation. If you’re looking to make an impact in the AI × Animals space, this is a concrete policy to build on.

*only if the underlying models are considered general-purpose AI with systemic risk (GPAISR), currently defined as being trained with 10^25 FLOP threshold

This means that AI systems affecting animals through agriculture optimization, synthetic biology, wildlife management, automated decision-making, etc* can now fall within regulatory scope if their actions create welfare harms.

Under its Safety & Security chapter, 𝗿𝗶𝘀𝗸𝘀 𝘁𝗼 “𝗻𝗼𝗻-𝗵𝘂𝗺𝗮𝗻 𝘄𝗲𝗹𝗳𝗮𝗿𝗲” 𝗮𝗿𝗲 𝗳𝗼𝗿𝗺𝗮𝗹𝗹𝘆 𝗱𝗲𝘀𝗶𝗴𝗻𝗮𝘁𝗲𝗱 𝗮𝘀 𝗮 𝗰𝗮𝘁𝗲𝗴𝗼𝗿𝘆 𝘁𝗵𝗮𝘁 𝘀𝗶𝗴𝗻𝗮𝘁𝗼𝗿𝗶𝗲𝘀 𝗺𝘂𝘀𝘁 𝗰𝗼𝗻𝘀𝗶𝗱𝗲𝗿 when developing general-purpose AI with systemic risk.

The week left us more connected as a team and clearer on the work ahead 🤝. A special thanks as well to Sanu — not pictured in the team photo 📸, but an important part of making this possible.

Thank you again to the CARE and Anima teams, and to everyone who joined us in these conversations.

We also had the chance to host:
• An 𝗔𝗜 𝗖𝗿𝗮𝘀𝗵 𝗖𝗼𝘂𝗿𝘀𝗲 with Damin and Jay 💡
• An 𝗔𝗜 𝗮𝗻𝗱 𝗟𝗼𝗻𝗴𝘁𝗲𝗿𝗺𝗶𝘀𝗺 meetup with Sam and Constance 🌌
• An 𝗦-𝗿𝗶𝘀𝗸 meetup with Sam ⚠️

Nineteen advocates role-played how animal agriculture might adapt if AI began shifting incentives of global powers 🌍 — from precision farming in China 🇨🇳 to new geopolitical dynamics. We’ll be sharing more soon.