OpenAI의 Sol에는 AI 안전 가드레일이 없다. 그 이유를 아무도 모른다

openai and anthropic have secured government clearance for their frontier models, sol and fable, despite a complete lack of public clarity on the specific safety thresholds or evaluation methodologies required. sam altman cited conversations with commerce secretary howard lutnick, but openai refused to detail the process, relying instead on third-party audits from groups like the uk aisi. the trump administration's executive order explicitly ruled out an fda for ai, leaving oversight fragmented across six cabinet agencies. this ad hoc system, compounded by reports of openai's political donations, creates an environment where access to powerful models is determined by unclear backchannels rather than established technical safety standards.