Anthropic has a new update for its coding tool, Claude Code. This update gives the AI more power to do tasks by itself. It also keeps safety rules to reduce risks.
Towards Self-Coding Without Constant Checks
This update comes as tech companies want tools that can work without user approval for every step. They want to balance speed and control.
The new feature is called "Auto Mode." It is in a research preview. It lets the system make decisions on running commands itself. Developers do not need to intervene for every step. TechCrunch reported this news. العربيه Business also saw the report.
Smart Safety System
"Auto Mode" uses a safety layer. This layer checks every action before it runs. It looks for bad behavior or unintended commands. This includes attacks like "Prompt Injection." Prompt Injection tries to trick AI into running harmful instructions.
If an action is safe, it runs automatically. Suspicious actions are blocked.
A Step Ahead of Rivals
This move follows similar self-coding tools. Companies like GitHub and OpenAI offer tools. These tools let AI do tasks for developers.
Anthropic's new idea is that the AI decides when to ask for permission. This gives it more freedom. But, there is still a check system.
Unclear Safety Standards
The company announced the feature. But they did not share details. They did not explain the rules for safe versus risky actions. Developers may have questions before using it widely.
A Full Set of Tools
This update follows other new tools. There is "Claude Code Review" for checking code. There is also "Dispatch" for assigning tasks to AI agents.
"Auto Mode" will soon be available for businesses and API users. It works now with Claude Sonnet 4.6 and Opus 4.6 models.
Advice to Reduce Risks
The company suggests using the new feature in safe places. Use it away from live systems. This can reduce possible harm if errors happen.
This shows a new stage for AI tools. They are moving from helping to doing tasks. Safety and trust are big challenges.