The trouble with AI is that it is being used to raise revenue and cheat people. Fines fines and fines . Flock cameras. Building and construction permits gone crazy. Saturation commercial's. Never mind wholesale privacy brokers, because it is safe to assume your data has already been stolen and misused.
Middle AI evils are limits on age, Election Ad's, disparaging politicians and little painters against AI generated parody. The is evidence of mainstream media whitewashing human rights violations.
Harmful AI is mining social media selectivity and using AI to make unreviewed judgements, including workplace performance (and trying to cut off any human appeal process - that is no natural justice at law).
Negligent AI is not using AI to tar and feather shonk contractors that are responsible for building defects galore, including on the spot fines from digital photographs (funny how traffic infringement's are the opposite).
Fake AI is using AI to describe expert systems, and deem prototypes as 'finished', and otherwise 'trusted' or deemed 'authoritative' with no audit, ever. A chatbot is not AI but a predictive LLM.