As artificial intelligence becomes capable of handling increasingly complex tasks, the question is no longer just what AI can do, but how it should be used. Anthropic is addressing that challenge with an updated usage policy for Claude, tightening and clarifying rules around political campaigns, weapons, surveillance and decisions that can affect people’s lives.

AdvertisementThe Puranic

Announced on October 8, the update also introduces a restriction on extreme cases of sustained, needless abuse directed at its AI models. The revised policy is scheduled to take effect on November 12, 2026.

One of the key changes concerns political activity. Anthropic prohibits the use of Claude to run deceptive influence campaigns, create fake news websites, impersonate candidates or election officials, and spread misinformation intended to mislead voters or suppress turnout. However, the company has removed its blanket ban on personalised voter targeting, allowing legitimate activities such as preparing election information in different languages. Deceptive targeting and misuse of personal data remain prohibited.

The policy also draws clearer boundaries around weapons and surveillance. Users cannot employ Claude to help arm drones, develop weapon-enabling systems or track people without consent. Using the AI to recommend whom law enforcement should investigate or charge is also prohibited, although legitimate work such as journalism, legal research and fraud monitoring remains allowed.

Anthropic is placing particular emphasis on human oversight when AI supports decisions involving healthcare, finances, legal rights, employment or essential services. The company says qualified people must remain involved in these consequential decisions.

The update reflects a broader challenge for AI developers: expanding what their products can accomplish while reducing the risks of misuse. As AI tools become more independent, clear rules and meaningful human accountability are becoming increasingly important.