Project summary

My paper on using access controls to address the dual-use nature of AI models got accepted to the ICML 2025 technical AI governance workshop, and I retroactively seek funding to cover the travel costs of going there.

I decided to go to ICML even if it stretched my finances thin because I was the sole author of the paper, so if I didn't go, the work wouldn't be presented.

What are this project's goals? How will you achieve them?

I had two main goals.

The paper is about the dual-use nature of AI models (how the same model can be used for good or harmful purposes) and why current approaches don't properly control this. In the paper, I propose verification-based access controls as one potential solution, together with a possible technical implementation. The work is still pretty conceptual and needs many holes to be filled in by future work, but I do think something like access controls will eventually be part of AI systems, and maybe this paper can help start that conversation.

My second goal was to go to ICML for the networking and learning. I'm mid-transition from technical AI safety to technical AI governance, and it was illuminating to talk to other people in the field at the workshop — some of the conversations helped me frame my plans for the PhD I'm now starting! This was my first conference, and I think it was quite a significant milestone for my career.

How will this funding be used?

Actual numbers:

• Student conference ticket: $550

• Flight from Prague and back: $1 200

• Accommodation: $550

I've gotten a grant of $1000 from the workshop itself, hence I'm seeking the remaining $1300. Since this was a solo author paper, I can't get support from my lab at ETH Zurich.

How much money have you raised in the last 12 months, and from where?

Last summer, I was part of Team Shard, which got compute funding from Neel Nanda/Manifund and general support funding from LTFF for our MATS project on gradient routing.