DeepMind urges culture-aware AI tests as English bias fuels blind spots
DeepMind is urging an overhaul of English-first benchmarks. The aim is for AI to respect local norms and avoid cultural missteps.
Read the original →Article digest · Oct 6 PM
AI Ethics — what each story says
Every story in this issue with a short summary, so you can decide what to read before leaving the site. Summaries are written by AI (Claude) only from each publisher's own description and checked by Jev for anything not in that description. They are not full translations. Follow "Read the original" for the full story.
9 stories8 with summaries5 topics
⚖️ Bias, discrimination and fairness2
- Headlines only (the publisher did not provide a description)
- When AI gets it wrong, real people pay the priceHannah Ferguson | Substack · 10/05 20:01 UTC
🔒 Privacy and surveillance1
How Europe's 'Third Way' on AI Justifies Turning Citizens Into Data Sacrifices
A new study of Hamburg's AI surveillance trial examines Europe's 'third way' compromise. It says the approach absorbs civic criticism while legitimizing the extraction of citizen data for machine use.
Read the original →🎨 Creators, copyright and consent3
Sony Music Seeks Removal of More Than 260,000 AI Deepfakes Impersonating Its Artists
The number has nearly doubled since March. Unauthorized AI tracks imitate artists including Harry Styles, Adele, Britney Spears, Queen and Michael Jackson.
Read the original →Are you part of the long tail? Why we should all be part of Australia's AI training data conversation
Copyright deals can protect professional creators, but much of AI's value also comes from the vast collective output of the open internet. Australia has an opportunity to recognise, protect and fund that shared data commons.
Read the original →CBA signs licence covering newspaper text in workplace AI
CBA is the first major Australian bank to sign a Copyright Agency Business Licence. The licence covers the use of Australian newspaper text in workplace gen AI tools.
Read the original →🎭 Misinformation and deepfakes1
ChatGPT users, your AI-written text will soon carry a hidden watermark
Text generated with ChatGPT could soon carry an invisible watermark. The watermark would make it easier to identify as AI-generated.
Read the original →🧒 Protecting children and vulnerable people2
When a Mental Health Chatbot Understands Your Culture — But Not Your Crisis
Making AI more culturally fluent could help people feel understood. But feeling understood and receiving safe mental-health guidance are not the same thing.
Read the original →American Academy of Pediatrics issues AI guidance for doctors
The American Academy of Pediatrics has issued guidance for doctors on the safe, equitable and effective use of generative AI in pediatric clinical care.
Read the original →The analysis and visual summary are in the Oct 6 PM issue. Japanese version here.