A hazard analysis framework for code synthesis large language models25 de julho, 2022 às 04:00OpenAI BlogVer notícia original