Have you ever used an AI chatbot and, because its answers didn’t meet your expectations, couldn’t help complaining—or even saying something rather rude? If so, you don’t really need to worry too much, because this kind of situation is not uncommon. However, the AI company Anthropic recently updated its chatbot Claude’s “Usage PolicyIt explicitly states that users must not persistently treat Claude maliciously or bully it. The new rules are expected to officially take effect starting November 12, and have also sparked discussions about the boundaries of human-AI interaction.
Anthropic’s new rule: Users, please don’t bully Claude.
However, this rule does not prohibit all expressions of criticism or dissatisfaction. Anthropic emphasizes that the policy targets the small number of users who repeatedly and without reason verbally attack or harass Claude over a long period. If users are conducting academic research, testing the model’s capabilities, writing story content with negative plotlines, or simply questioning and rebutting AI responses, these actions will not be considered violations. Therefore, ordinary users need not worry about being restricted because of a momentary emotional reaction or for offering sharp opinions.
In fact, Claude itself already has mechanisms for handling malicious interactions. When the system detects that a small number of users are persistently engaging in abusive, harassing, or bullying behavior, Claude can proactively end the conversation to prevent the situation from escalating further. Anthropic has also said that directly ending the conversation will remain the primary way to handle such issues, rather than immediately punishing users.
In addition to adding an anti-bullying clause, Anthropic’s updated usage policy also covers several important areas, with the goal of making the rules clearer and easier to understand. One key point is prohibiting the use of Claude to create fake accounts, fabricate news outlets, or spread misleading information. This restriction applies not only to commercial activities but equally to political advocacy and public issues.
In election-related applications, Anthropic has further tightened its restrictions. The new policy explicitly prohibits using Claude to deceive voters, impersonate candidates or election workers, or engage in any conduct that hinders people’s right to vote. At the same time, the company has adjusted some previously overly strict rules. For example, the previous blanket restriction on personalized political advocacy could affect public-interest uses such as nonprofits providing voting information in different languages, so the relevant provisions have been revised. However, using personal data for manipulative ad targeting, or influencing voters’ decisions through deceptive means, remains explicitly prohibited.
In the areas of security and law enforcement, Anthropic has also drawn clearer red lines. The updated policy states that Claude may not assist in developing weapons-related software or hardware, including uses such as integrating weapons into drone systems. In addition, AI may not be used to surveil individuals without their consent or to help law enforcement agencies determine specific targets to pursue. However, legitimate uses such as legal research, fraud investigations, risk assessments, and report analysis are still permitted.
For ordinary users, these policy adjustments actually will not have much impact. Most people can still use Claude as usual for work, study, or various everyday queries. Anthropic specifically points out that if AI is used in important decision-making areas such as health, finance, or law that may affect individuals’ rights and interests, there must be professionally qualified personnel who can review and correct the recommendations provided by AI, and the affected parties should also be informed that AI was involved in the decision-making process.
Overall, Anthropic’s new policy reflects that, as AI technology becomes increasingly widespread, platforms are placing greater emphasis on safety, transparency, and responsible use. For the vast majority of users, as long as they interact with AI reasonably and in good faith, these rules will not create an extra burden. In other words, unless someone deliberately treats bullying Claude as entertainment, the future user experience will largely remain the same as it is now.
Source: KOCPC Chinese