News from France sounds exactly backwards to me:

French intelligence officials to use older intelligence data, including data the government isn’t currently allowed to retain, to train AI systems.

Such an approach should be called out for what it is, repeating the worst mistakes in history at faster speed with less oversight.

Think of it this way, if you predicted any future police action in France from learning their past tragic history of colonialism you would repeat it instead of shifting towards what should happen instead.

I just recorded a new presentation for the 2021 RSA Conference about this exact problem. AI can’t be implemented as a detection system for terrorism without the heavy hand of human philosophy and control over what is defined as future terrorism.