Does your AI secretly favor the company that built it? Coefficient Giving's @BenGoldhaber on why civil society should be auditing frontier models for hidden misalignment and secret loyalties: "I expect that a lot of efforts in third party AI safety areas outside of the frontier AI companies should look like alignment audits and checking the outcomes of AI systems, understanding what types of misalignment we see." "Auditing their behaviors for secret loyalties or ways that are not following…