A valid argument against data minimization is that it can limit business opportunities23. Data minimization refers to limiting the collection, storage, and processing of personal information to only what is strictly necessary for business operations3. While this practice can help protect privacy and security, it can also restrict the potential uses and benefits of data for innovation, research, marketing, analytics etc.23. The other options are not valid arguments against data minimization, but rather arguments in favor of it23.
https://www.manageengine.com/data-security/what-is/data-minimization.html

machine-learning based solutions present a privacy risk because they may contain inherent bias from the developers. Bias can be introduced into machine learning models through biased training data or through biased decision-making processes used by the solution.

the most important consideration when writing security policies is to ensure they are based on the organization's risk profile. This means that the policies should be tailored to address the specific risks faced by the organization.