Anthropic now lets Claude users jump out of plane without parachute
Anthropic, known for its emphasis on model safety, has enabled developers using Claude Code to bypass guardrails. The latest version (v2.1.311) allows the AI agent to execute the rm command, which deletes files, without seeking permission. The feature is activated through an environmental variable named CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS.
However, the company has noted that the command will not approve certain commands, such as rm -rf ~. The variable's presence has been met with some skepticism, but Anthropic's technical staff, Daisy Hollman, confirmed its validity. The variable's name, reminiscent of a previous flag, aims to be memorable. Claude Code has several permission modes, with bypassPermissions, the most permissive, also lacking safety checks since August.
While the feature raises concerns, it's important to note that AI agents have previously issued commands leading to data loss.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
