Claude Code Auto Mode Becomes Default

Anthropic will apply Auto mode as the default setting for new sessions on Pro, Max, and Team plans in Claude Code starting August 14.

In an evaluation commissioned by Anthropic, 1,053 paid developers were tested on their approval of dangerous commands:

• Rate at which human reviewers rejected dangerous tasks: 13.6%
• Rate at which Auto mode blocked dangerous tasks: 89%

Another third-party evaluation firm, Trajectory Labs, assessed the latest public versions of Claude Code and Codex as of July 17, 2026, targeting 72 indirect prompt injection scenarios with a total of 720 attacks. According to Anthropic, Claude Fable 5, Opus 5, and Sonnet 5 did not allow any attacks to succeed in Auto mode.

However, these results are based on evaluations published by Anthropic, and additional independent verification is needed to determine whether they provide sufficient defense against supply chain attacks disguised as trusted commands, such as malicious third-party packages.