Theo Browne discusses the safeguardsAI safety is the field and practice of reducing harmful failures, misuse, loss of control, and unintended consequences from AI systems. around Anthropic’s Fable model, including routing certain requests to another modelAI model routing sends each request to a model chosen for that request's complexity, cost, speed, privacy, or other requirements. and the potential for benign requests to trigger restrictions.
Theo Browne examines data retention termsData privacy governs how personal, confidential, or sensitive information is collected, used, shared, retained, and protected in AI systems. and criticizes initially undisclosed restrictions on frontier model development tasks. He describes a later change to make such safeguards visible and argues that hidden interventions create a trust problem for developers. His explanation of the company’s motives is explicitly speculative.
Watch on YouTube
