“The argument against open-source AI is often based on the existential risk that bad actors could use the technology for malicious purposes.”