Anthropic updates Claude rules, allows disconnect and restricts deceptive campaigns
Technology
Anthropic just rolled out new rules for using its AI, Claude, after more than a year.
Now users can't be needlessly harsh or abusive toward the AI, and if things get out of hand, Claude can actually disconnect from those users.
The policy also cracks down on using Claude for deceptive commercial or political campaigns, especially around elections.
Claude used in South Korea hacks
These changes come as a CrowdStrike report revealed that Claude was used in hacks targeting South Korean banks.
The updated rules also ban using the AI for weapons control in projects like drones or self-driving cars and made its surveillance restrictions clearer.
Anthropic says these steps are all about keeping people (and the AI itself) safer and stopping future high-risk misuse.