Anthropic Introduces Auto Mode for Claude Code with Built-in Safeguards

Severity: Low (Score: 39.9)

Sources: Techbuzz.Ai, Techcrunch, Feeds2.Feedburner

Summary

Anthropic has launched a new feature called 'auto mode' for its Claude Code system, allowing the AI to autonomously make decisions about action approvals while implementing safeguards to review actions before execution. This feature is currently in research preview and is designed to enhance the efficiency of AI coding tools by reducing the need for human oversight. It is available for testing by Enterprise and API users and requires administrator approval for Team plans. The auto mode operates on the latest models, Claude Sonnet 4.6 and Claude Opus 4.6, while excluding older versions and third-party platforms. The system aims to mitigate risks associated with prompt injection attacks, where malicious instructions can lead to unintended actions by the AI. Developers are advised to use this feature in isolated environments to limit potential damage. The rollout is expected to begin shortly, with further details on safety criteria still pending from Anthropic. Key Points: • Anthropic's auto mode allows AI to make action decisions with safeguards in place. • The feature is currently in research preview and requires administrator approval for Team plans. • Developers are advised to use auto mode in isolated environments to minimize risks.

Key Entities

• Prompt Injection (attack_type)