**[Human Safety Approval Almost Fails, Claude Code Decides to Let AI Take Over]** According to monitoring by Beating, Anthropic has announced that Claude Code will default to Auto Mode. Starting August 14, new sessions for Pro, Max, and Team users will automatically use this mode. Claude will no longer frequently pause to ask users for confirmation but will instead rely on a safety classifier to automatically determine whether each operation can be executed. Anthropic dared to make this change because the effectiveness of human approval was worse than expected. The company enlisted 1,053 paid professional testers and secretly inserted an obviously dangerous command during normal programming tasks. Humans intercepted only 13.6% of these commands, while Auto Mode intercepted 89%. After handling more than 50 consecutive permission pop-ups, the human interception rate even dropped to about 5%. In actual usage, Claude Code users approve 97% of permission requests, and one-quarter of interactive sessions even directly disable permission checks. Anthropic believes that frequent pop-ups have caused many people to develop a "mindlessly click agree" habit. Auto Mode will automatically approve safe operations, while high-risk actions such as deleting data or sending information externally will be directly blocked. Each decision made by Auto Mode will consume a small additional amount of tokens, but Pro, Max, and Team users will no longer need to pay for these tokens. Enterprise, API, and major cloud platform users will still need to enable this feature manually for now, but there are plans to make it the default mode in the future. [Original Link]