Who Gets to Decide A.I.’s Moral Code?

Companies developing large language models make decisions about which values and priorities their systems will reflect, and these design choices affect how millions of users interact with the technology. OpenAI, Google, Anthropic, and Meta have each implemented different approaches to how their AI systems respond to queries about sensitive topics. Researchers including those at Stanford University and UC Berkeley have documented how these systems produce different outputs on identical questions depending on their underlying training data and safety guidelines. No formal regulatory framework currently governs these decisions, meaning companies set their own standards.

Next story