Surveillance
Correct systems can still be wrong.
Facial recognition was treated both as a technical risk when it fails and as a civil-liberties risk when it works exactly as designed.
OCOM5202M / Ethics of AI
Ethics of AI
The assessed work covered discussion posts and a scripted presentation exploring facial recognition, mass surveillance, copyright, generative AI, public policy and the difference between technical capability and social permission.
I have not reproduced third-party images or teaching material here. The useful portfolio point is the thinking: AI ethics is not an afterword to engineering, it is part of deciding what should be built, deployed or resisted.
Generative AI
Data has politics
The copyright and training-data work pushed beyond “can we train this?” into consent, attribution, economic displacement and control.
Policy
Tech ships before the rules do.
The module made regulation feel less abstract: once systems are embedded, changing social defaults is much harder.
Lesson
Engineering needs a conscience
The practical lesson was to ask who benefits, who is exposed to risk and what power asymmetries the system amplifies.