Anthropic will enable auto mode as the default setting in its AI coding agent Claude Code starting August 14. This change aims to solve the approval fatigue problem, since 93% of requests were being manually approved under the old system. Auto mode uses a two-stage classifier model to evaluate risk before executing actions. Low-risk actions like reading a file execute automatically, while higher-risk operations still require manual approval. The feature, first launched as a research preview on March 24-25, 2026, works across Anthropic API, AWS Bedrock, and Google Vertex.
Source: Read the original article
