Meta's move on AI bias raises risk, eyebrows

Meta has unveiled Llama 4, its latest AI model, aiming to tackle historical left-leaning biases in AI outputs. The company asserts that this initiative addresses concerns about AI neutrality in social and political contexts. However, critics suggest that the effort may be more about appeasing conservative critiques than ensuring genuine neutrality. Experts highlight the technical difficulties and subjectivity involved in modifying large models, cautioning that such adjustments could inadvertently introduce new biases. Human rights organizations express alarm over reports that Llama 4 references harmful and debunked practices, such as conversion therapy, raising ethical questions about the model's content and the broader implications of bias mitigation in AI development.