Allow me to clarify why I stated "incorrect data is worse than no data"

I had in mind AlphaGo Zero beating AlphaGo, see: https://www.scientificamerican.com/article/ai-versus-ai-self-taught-alphago-zero-vanquishes-its-predecessor/
This means that no data is sometimes even better than having data at all, even correct data. While wrong data, can lure your company in the wrong direction, thinking your product is performing or your financials are good, until you realize there was a quality problem, and maybe that’s too late. At least with no data, you know you have nothing, so the human mind or the AI algorithm will teach itself.
Regarding your point, I agree: when it comes to data quality, you should always assume the signal is noisy and you need to build defense against that.