According to CoinMeta, Anthropic announced that Claude Code will default to automatic mode starting from August 14th. New sessions for users of pro, max, and team will automatically use this mode. In this mode, Claude no longer stops frequently for user confirmation; instead, a security classifier automatically determines whether each operation can be executed. Official tests showed that human approval had poor effectiveness, only blocking 13.6% of dangerous commands, while the automatic mode was able to block 89%. In actual use, Claude Code users approve 97% of permission requests, and the frequent pop-ups have led users to develop a habit of "mindlessly clicking agree." The automatic mode will automatically allow safe operations and directly intercept high-risk ones. Each judgment consumes a small amount of token, but users of pro, max, and team do not need to pay for this.