Weekly Links (11/21/25)

• by Elliot Lewis
  1. Why aren’t more people in the U.S. using AI models from China? There are many possible answers.
  2. “I think I’m deeply uncomfortable with these decisions being made by a few companies, by a few people,” Dario Amodei (CEO of Anthropic) told Anderson Cooper in an interview.
    COOPER: Who elected you and Sam Altman?
    AMODEI: No one! No one! Honestly, no one.
  3. Anthropic found that intentionally training models to "reward hack" causes the model to exhibit unrelated misaligned behavior.
  4. Character.AI is removing open-ended chats for users under 18 next week.
  5. Ahead of the holidays, consider the dangers of AI toys for children.