
Anthropic will make Auto mode the default for Claude Code's Pro, Max and Team users from August 14, reducing the need for manual approval at every step. The AI will now autonomously…
Anthropic will make Auto mode the default for Claude Code's Pro, Max and Team users from August 14, reducing the need for manual approval at every step. The AI will now autonomously continue coding tasks unless it detects an action that could be irreversible, destructive or outside the user's environment. Users can revert to manual mode anytime using a shortcut. The feature remains opt-in for Enterprise users for now.

Anthropic tested Auto mode with 1,053 paid users and claims its classifier caught 89% of dangerous commands, while human reviewers caught only 13.6%. The company also found that users clicked 'approve' on 97% of permission prompts, suggesting 'approval fatigue.' Among Teams and Enterprise users, those on Auto mode reportedly shipped 25% more pull requests. Customers including Adobe, Nuro, Gusto and Garner Health already use Auto mode as their default. Boris Cherny, Head of Claude Code, said the team has used Auto mode exclusively for months.
Claude Code's Auto Mode is being sold as a productivity booster, and the 25% more pull requests claim will be worth watching when independent teams validate it. But the narrative that AI safety classifiers (89% catch rate) easily outperform humans (13.6%) is the real spin. That comparison pits a focused algorithm against tired, approving-97%-of-prompts developers, a straw man if there ever was one. The more honest test: will the classifier catch truly novel or adversarial attacks that its training data never saw, or only the familiar dangerous commands it was built to spot? That question remains unanswered.
Sources (4): gadgets360.com, timesnownews.com, livemint.com, businesstoday.in
This story was synthesised by AI from the 4 sources linked above.
Updated: this story now draws on 4 sources.