Besiroglu says OpenAI did have access to many of the FrontierMath problems and solutions — but he added “we have a verbal agreement that these materials will not be used in model training.”
It’s not like the company building a plagiarism tool would use said tool to plagiarize training data. That would be inconceivable.
Kind of reminds me of the government funding Star wars programs that never produced anything but was credited for spending the Soviet Union into their grave because they couldn’t keep up. But I don’t think it’s going to work the same this time…