The problem is, as someone in the field of AI, I can see how much overhyped low-quality products are sold to stakeholders believing they are getting a human-like or godlike artificially intelligent thing, when it's just a bunch of linear regressions or multilayer perceptrons operating on unsanitized data without proper evaluations or misleading numbers in the marketing, careless ways of separating training and test data etc.
I think what the article says is likely to be true. It's a scientific cover-your-ass label for the high-ups to be able to do whatever they want and justify it. Similar to how CEOs bring in external consultants and pay them fat money so the CEOs can now do whatever they wanted originally, but now with Big Consulting Firm's stamp of approval. You can make these "AI" models skew to the direction you want. Just as it is with statistics and p-hacking, only that the machine learning community is even less versed in confidence values and is just generally less mature in terms of best practices as it's a newer field.
And absolutely researchers should pay attention to who they work for and where the money comes from. Fundamental research is one thing, working on general AI is useful for all human endeavors just like working on energy efficiency or better cars. It's a necessary part of life that bad actors have benefits from a tide raising all ships.
But it's not the same when directly working on a project for a bad actor. Similar to how Google engineers stood up against military projects or serving the Chinese censoring machine, it is important for mathematicians and computer scientists to consider whether they are building something unethical. These are professionals, intellectuals who have more responsibility than lower paid workers. We cannot blame the cleaning personnel or the cooks at the police canteen in the same way.
It is part of one's civic duty as an intellectual to reflect upon one's societal role and impact.
Again, this is not about fundamental research, that might be used for evil purposes. It's about directly working for corrupt organizations. Whether the police in general (or which specific levels or branches) is corrupt enough to refuse working for them and how much worse they are compared to big corporations is hard for me to judge from Europe. But it's certainly something that people working for them should reflect on and make a moral choice because they are the ones who see what they actually work on. There are many places for mathematicians and computer scientists to work at and so they have the luxury to follow their conscience. And all this also applies to intelligence agencies, like the NSA and others as well.