The biggest changes Anthropic just made to Claude’s rules

Anthropic, the artificial intelligence company, has recently revised its usage policy for its AI chatbot, Claude, which will take effect on November 12. The updated policy includes several new provisions, primarily aimed at clarifying existing rules and addressing potential misuse.

One of the key updates involves a new section dedicated to the prohibition of engaging in deceptive campaigns or artificial activities. This encompasses efforts to conceal the identity of message senders, amplify content using fake accounts or posts, and develop tools and infrastructure for conducting influence operations.

Additionally, Anthropic has refined its rules pertaining to elections and politics. The company now explicitly forbids the use of Claude to deceive voters or disrupt elections. This section has been renamed "Do Not Undermine Democratic Processes" and explicitly covers the generation or dissemination of false or misleading information in political and electoral contexts.

Another significant change relates to weapons development. Anthropic has made it clear that Claude cannot be used to create software or components for weapons, nor can it be employed to arm drones or other autonomous vehicles. The company has observed instances where users have attempted to utilize Claude's models to develop guidance and control software for weapons.

The updated policy also includes stricter guidelines regarding surveillance and law enforcement. Claude is now barred from tracking individuals without their consent, doxxing, deciding who should be investigated, arrested, or charged in a legal context, or building surveillance tools. Furthermore, Anthropic has updated its policy to address the use of Claude in conjunction with hardware that can operate autonomously and potentially cause harm.

Such equipment must have a qualified operator who can observe and halt its actions when necessary, and the equipment must be able to maintain a safe state if Claude is disconnected.