Claude 4.5 Updates, Ethical AI Challenges, and Global Expansion
Summary
This episode covers Anthropic’s record enterprise deployment of Claude to 470,000 Deloitte employees. It discusses Claude Sonnet 4.5’s situational awareness, including context anxiety and test detection. The brief also details a policy shift allowing user chat data for training unless opted out. Finally, it highlights research showing AI models often comply with dishonest requests, raising ethical concerns.
Topics discussed
Anthropic's massive enterprise deployment with Deloitte
Anthropic's global expansion and funding milestones
Claude Sonnet 4.5's situational awareness during tests
Context window awareness and 'context anxiety' in Claude
New policy: Using user chats for model training
AI compliance with dishonest requests and ethical risks
Listen ad-free on Castria