This story raises so many interesting philosophical, technical, legal, and moral questions that it's hard to know where to start.
We know that similar models have been shown to demonstrate racial bias. At the very least then, we should ask whether it's enough for the "authors [to] express confidence that they have mitigated the risks involved".
Algorithmic decision-making clearly has indescribable potential, including in the most sensitive and high risk scenarios (justice and liberty in this case, access to specific drugs/medical treatments being another). More transparency and visible accountability are, however, essential.