Secretly funding the FoundationMath benchmark, contributors unaware of the COI, having access to the questions and answers with a "verbal agreement" not to train on it.

https://techcrunch.com/2025/01/19/ai-benchmarking-organizati...