Then those same organizations that "give in" end up providing inferior AI tools, and you face the dilemma of either using poor models and accepting poor results or sharing data with better models that may not be governed correctly.