From hard refusals to safe-completions: toward output-centric safety training
How Amgen uses GPT-5
How Cursor uses GPT-5
GPT-5 System Card
Vision Language Model Alignment in TRL ⚡️
Stop falling for Hamas' propaganda and lies
Paradigm Shifts and the Winner's Curse
Providing ChatGPT to the Entire U.S. Federal Workforce
Estimating worst case frontier risks of open weight LLMs
Introducing gpt-oss