OpenAI said it fired three safety researchers for a “significant breach of trust,” and said the decision was not because they raised concerns about the company’s AI safety work.