Anthropic switches Claude Code to automatic mode by default: efficiency has surpassed manual review

• Auto mode will become the standard for Pro, Max, and Team plans starting August 14
• Tests: the automatic mode identified 89% of errors compared to 13.6% manually
• 97% of permission requests are automatically approved by users
• Anthropic adds protection against prompt injection and blocking rules

Anthropic announced that starting August 14, the automatic mode (Auto mode) will become standard for Claude Code users subscribed to Pro, Max, and Team plans. The system will now perform most actions without requesting permission, except for operations that may be irreversible or go beyond the user's working environment.

The first experiment with Auto mode was conducted in March of this year. At that time, it was positioned as a compromise between increased encoding speed and developer control. According to TechCrunch, in this mode, Claude Code continues to operate without additional confirmations, unless the action pertains to destructive operations.

During extensive testing involving 1,053 users of paid plans, the automatic mode demonstrated an unexpectedly high level of reliability. It was able to identify 89% of potentially erroneous actions, while traditional manual checks only achieved a rate of 13.6%. Researchers attribute the low effectiveness of manual control to users' habit of automatically approving requests: 97% of permission requests in the system are approved 'automatically'.

Boris Cherniy, head of the Claude Code division, noted in his post on the social network X that his team has been working exclusively in automatic mode for several months and sees no reason to return to confirmation requests. 'My team and I have been using only automatic mode for many months. I can't imagine going back to permission requests!', he wrote.

In addition to improving efficiency, Anthropic continues to enhance the security of Claude Code. The company's plans include implementing protection mechanisms against prompt injection attacks, as well as the ability to set customizable prohibition rules aimed at preventing data leaks.

Automating coding processes is important in the context of growing competition in the generative AI market. Claude Code, as one of the leading tools, aims to combine speed with reliable quality control, which is especially relevant for developers working on mission-critical projects.

What will change for users?

Pro, Max, and Team plan users will gain access to Auto mode without the need to manually confirm each action. This will reduce development time, decrease the number of human errors, and allow for a focus on more complex tasks. At the same time, Anthropic promises to maintain strict limitations on potentially dangerous operations.

Thus, the transition to automatic mode in Claude Code represents an attempt to combine the efficiency of AI-generated code with enhanced security measures, meeting the demands of both professional developers and companies seeking a reliable and fast development process.

Source: 3DNews