Anthropic says that starting August 14, "auto mode" will be the default setting in Claude Code, its AI coding tool, for customers on Pro, Max, and Team plans.
In practice, that means the software will take more actions on a developer's machine without pausing to ask for a human sign-off on each step. As TechCrunch put it, programming with Claude Code "will soon require even less human oversight."
The company's argument, according to a Techmeme summary of a post by developer and blogger Simon Willison, is that the tool is now good enough at catching harmful actions on its own to justify the switch. Willison noted the change was among the topics discussed in a fireside chat with Anthropic's Cat Wu and Thariq Shihipar.
One number is doing a lot of work in Anthropic's defense of the decision. According to egamers.io, the company is pointing to a prompt injection result of "zero out of 720" — a test of whether attackers could smuggle malicious instructions past the system — to back a default change the outlet says "will reshape how millions of developers work."
There's a second, less flattering rationale in circulation. The New Stack framed the shift bluntly: auto mode is becoming the default "because humans can't be trusted." Coverage from egamers.io points the same direction, describing Anthropic as targeting "rubber-stamped approvals" — the well-known habit of clicking "yes" on every prompt without reading it. Other outlets, including Pluang and 9to5Mac, have relayed the change as a safety-and-efficiency move landing next week.
Why it matters: this is a mainstream AI company deciding that its software supervises itself better than its users do, and quietly moving the default for paying customers to match — a template other AI tools are likely to copy.