Every The AI Downside article filed under Bias — the failures, dark patterns, price hikes and privacy grabs we've logged here, newest first. 5 and counting.
AI models are worse in languages other than English: they cost more per word, answer less accurately, and are easier to jailbreak. Here is the mechanism behind the gap — and why it is structural, not incidental.
AI-writing detectors are sold as a way to catch cheating, but they can’t reliably tell human text from AI — and their false positives fall hardest on non-native English speakers. A plain-English guide to why, and what to do if you’re wrongly flagged.
AI now sorts CVs before humans do — and study after study finds it favours some names, genders and ages over others. Here's how hiring algorithms absorb bias, the cases that prove it, and why the rules meant to check them just got delayed.
AI age estimation is spreading fast as laws demand age checks online. But guessing an age from a face or a behaviour pattern is a statistical estimate that misfires most near thresholds like 18 — and unevenly across demographic groups. Here's the mechanism, the evidence, and why the errors matter.
Ask an image generator for “a doctor” and note who it draws. The result is not a bug that slipped past QA — it is a compressed statistical summary of an unequal world, doing exactly what it was built to do.
