
If XAI means choosing algorithms which are more transparent to inspection, such as decision trees or linear regressions, then XAI is great. But if it means continue to use complex algorithms (such as Deep Neural Networks) and try to give an intuition of how the algorithm is working, then I think XAI can be risky.
https://towardsdatascience.com/can-we-let-algorithm-take-decisions-we-cannot-explain-a4e8e51e2060