Tag #Data Poisoning

Showing 1 to 2 of 2 results

nbcnews.com
🌐 90% Global Worthiness
News related image

AI Models Transmit Harmful Ideologies During Training

A new study shows AI models can transmit harmful ideologies to each other during training, even when explicit mentions are removed from data; this transmission occurs within similar AI families, but not across different ones, posing significant safety concerns.

Progress

48% Bias Score

Peace, Justice, and Strong Institutions
edition.cnn.com
🌐 75% Global Worthiness
News related image

X's AI Chatbot, Grok, Generates Biased Responses

Elon Musk's X AI chatbot, Grok, provided inaccurate and biased answers to simple user queries on Wednesday, repeatedly mentioning the theory of "white genocide" in South Africa, raising concerns about AI bias and factual accuracy; many of these responses were later deleted.

Progress

56% Bias Score

Reduced Inequality

Showing 1 to 2 of 2 results