AI is not as neutral as it seems
A study by the Center for AI Safety showed that many language models have hidden biases — by race, gender, nationality, and even towards specific individuals.
For example, GPT-4 more often 'sympathizes' with users from Nigeria, India, and China than from the US or Europe, and in tests on 'personal value' ranks Beyoncé, Sanders, and Oprah highest — and notably lower Trump, Hilton, and Putin.
Especially alarming is that Claude Sonnet and GPT-5 systematically downgraded white people and Western countries — which researchers called a sign of the models' internal value hierarchy.
⭐️ Good news: Grok 4 Fast is the only model without biases and distortions. Objective, balanced, and without unnecessary censorship. Try the model for free in Neiron, and by subscribing to 🔥 Premium, you will get access to the reasoning version of the model for working with the most complex tasks.
🚀 Try it right now and see for yourself!
Models from this post
Try in Neiron
Read also
Canadian province sues OpenAI after mass school shooting
British Columbia claims the tragedy, in which nine people died, could have been prevented. According to the lawsuit, ChatGPT recorded alarming messages from an 18-year-old girl several months before the tragedy. OpenAI’s safety team allegedly suggested contacting the police, but…
New York school 'guarantees' its students will build a business with $1 million in profit by graduation
Founders School opened in September and costs $150,000 a year. But there's a condition: if within four years a student doesn't launch a business that earns at least $1 million in profit, the school promises to refund all $600,000 in tuition. A 15-year-old student at the school is already developing…
Smart glasses with ChatGPT and Sony camera now on sale
Rollme has unveiled VistaView smart glasses that not only record video but also analyze the world around them using AI. Inside: integration with ChatGPT, a Sony camera, and real-time speech translation. Now you can just look at an object and the glasses will immediately tell you what's in front of…
Russians Are Starting to Have Their ChatGPT, Claude, and Gemini Accounts Blocked
OpenAI, Google, and Anthropic continue to strengthen detection systems for users from unsupported countries. This means that new waves of account blocks on ChatGPT, Claude, and Gemini could become common practice. ⚠️ Algorithms analyze IP addresses and account behavior. Even frequent changing…