Anthropic bans systematic abuse of Claude and election interference

Restrictions on abusive behaviour towards AI

Anthropic has updated the terms of use for its artificial intelligence systems for the first time in the past year. Among the key changes is a new ban on “systematic and nonsensical abusive or cruel behaviour” towards the Claude model. According to the updated regulations, if a user begins to use abusive language, the AI will terminate the dialogue. However, the company has not disclosed whether additional sanctions, such as blocking the accounts of violators, are envisaged.

Anthropic emphasised that this rule does not apply to ordinary frustration, discussions with the model, writing dark creative plots, system testing, or scientific research. The restrictions will apply exclusively in “extreme cases”.

Ban on interference in political processes

In addition to ethical norms, the company introduced strict restrictions on the use of Claude to interfere in electoral processes. It is prohibited to spread disinformation about candidates or voting procedures, impersonate candidates or election commission staff, and make any attempts to reduce voter turnout.

Restrictions on surveillance and security

The new rules also provide for restrictions on the use of Claude for weapons development, as well as in the fields of healthcare and finance. Particular attention is paid to the ban on surveilling people without their consent, specifically political dissidents, and the use of AI to make decisions regarding arrests.

All these innovations will come into effect from 12 November.