John Jumper’s switch from DeepMind to Anthropic could reshape how we build safe AI models, highlighting new approaches to responsible AI development and co