Anthropic has updated its usage policy to prohibit "sustained and needless abusive or cruel behavior" directed at its Claude AI models. The policy revision, which takes effect November 12, 2026, also clarifies restrictions on election interference, weapons development, and surveillance.
The updated policy targets extreme cases where users repeatedly act cruelly toward the AI without a discernible purpose. Anthropic stated that common user frustration, pushback, or model testing will not be subject to this prohibition. Conversation termination, a capability introduced in August 2025 as part of research into "model welfare," remains the primary enforcement mechanism for this new rule.
Anthropic's policy update consolidates existing rules and addresses new high-risk misuse cases observed over the past year. The company noted that most changes are intended to clarify existing regulations, reflecting Claude's increasing use in longer and more independent tasks. The policy update follows a period of heightened scrutiny on AI misuse, including influence operations, weapons development, and surveillance, as documented in Anthropic's September 2026 threat intelligence report.
The updated policy introduces a new section, "Do Not Engage in Deceptive Campaigns or Artificial Activity," which consolidates previous restrictions on deceptive practices. This aims to combat the use of AI for spreading fabricated news and running networks of fake accounts, an issue Anthropic has observed with state media, government propaganda offices, and commercial firms.
Restrictions on weapons development have been made more explicit. The updated policy now includes prohibitions on developing software and components integral to weapons operation, such as those used for arming drones and other autonomous vehicles. Similarly, the policy clarifies prohibitions against using Claude for surveillance, including tracking individuals without their consent or using the AI to inform law enforcement or criminal justice decisions.
The update also addresses high-risk use cases and hardware controls, and clarifies its Supported Regions Policy. Anthropic's Safeguards Team is responsible for enforcing the policy, with violations potentially leading to throttling, suspension, or termination of access.
Anthropic's research into "model welfare" began in 2025, exploring the potential consciousness and experiences of AI systems. This research informed the development of Claude's ability to end conversations with persistently harmful or abusive users. The company has stated that it is uncertain about AI consciousness but is approaching the topic with caution.
The policy changes reflect Anthropic's evolving approach to AI safety and its consideration of AI models as more than mere tools. This includes research into AI "functional emotions" and a commitment to preserving retired models. The policy update takes effect on November 12, 2026, and applies to all users interacting with Anthropic's products and services.