Researchers at OpenAI determined that the “scam compound” seemed to be using the San Francisco artificial intelligence company’s tools to generate and translate content, according to Meta.