Abstract:
Anthropic recently released a new usage policy applicable to Claude AI, which will take effect on November 12, 2026. The new version of the policy mainly adds clauses prohibiting unwarranted abuse of Claude, and the relevant clauses applicable to elections, weapons development, surveillance, etc. are also changed simultaneously.
No unwarranted insults or abuse of Claude:
A agency states that continuous and unnecessary abuse or cruelty to models is a clear prohibited behavior. This regulation targets behaviors that are repeated, extreme and with no apparent legitimate purpose. However, users expressing dissatisfaction with model answers, proposing refutations, creating dark themes, or conducting model testing and research purposes are not prohibited (of course, the right of interpretation belongs to A agency).
Claude has previously had the ability to proactively end conversations where abuse continues. Agency A said that this is still the main way to deal with such behavior, that is, the ended conversation cannot continue to send messages, but this will not affect other chats in the user's account, and certainly will not result in a ban.
Simultaneous adjustments to election, weapons and surveillance provisions:
The new version of the usage policy concentrates restrictions on false propaganda and manipulation into deceptive activities clauses, covering the use of fake accounts, fictitious media or websites to disguise the source of information, creating false buzz, and building related tools and infrastructure. The scope of application includes both political and commercial activities.
Election provisions focus on deceiving voters, suppressing votes or interfering with elections. Company A has also removed restrictions on personalized voting and campaign targeting, but related operations using deception or abuse of personal data are still prohibited.
The weapons terms specifically prohibit the use of Claude AI to develop or operate software and components required for weapons, equipping platforms such as drones with weapons, and activities that increase the lethality or spread of biological or chemical agents.
The surveillance and enforcement provisions prohibit tracking individuals without consent, having Claude determine or recommend targets for investigation and arrest, or developing surveillance tools; tracking, content moderation, news reporting, and legal research with the person's consent are still allowed.
Comments