OpenAI said Thursday that it has seen several foreign influence campaigns tap the power of its AI models to help generate and translate content, but has yet to see novel attacks enabled through its tools.