Data poisoning could turn open-weight models into censorship-resistant coordination tools
Great post! I’ve also been interested in what data poisoning for hidden loyalties means for middle powers
that's a really interesting question! I wonder if forethought is doing any research here?
Great post! I’ve also been interested in what data poisoning for hidden loyalties means for middle powers
that's a really interesting question! I wonder if forethought is doing any research here?