My current guess (as a relatively uninformed outsider) is that actors with some combination of pre-existing domain knowledge, operational access, and (obviously) intent are most concerning.
Re defender-first access: yes definitely seems like a threat (e.g. misuse by insiders, account compromise/coercion, leakage of outputs and workflows, etc). I don't really have more concrete thoughts beyond this though. It would be useful to see a detailed estimate (if there isn't already one internally) on how robust Rosalind Biodefense and Project Glasswing are to this.
Super exciting!
Congrats Jordan, best of luck! :)
So cool that you’re working on this course! Excited to learn more once its released.
Really valuable work - thank you. Looking forward to the Animal Welfare Alignment Newsletter!
My current guess (as a relatively uninformed outsider) is that actors with some combination of pre-existing domain knowledge, operational access, and (obviously) intent are most concerning.
Re defender-first access: yes definitely seems like a threat (e.g. misuse by insiders, account compromise/coercion, leakage of outputs and workflows, etc). I don't really have more concrete thoughts beyond this though. It would be useful to see a detailed estimate (if there isn't already one internally) on how robust Rosalind Biodefense and Project Glasswing are to this.
Thanks :)
Great post!
Thanks for your comment.
I’m not able to follow what you mean here, could you explain your points more simply?