Anthropic's updated usage policy bans tracking people without consent
The 2026 Usage Policy adds rules on deceptive campaigns, surveillance and physical hardware safety, and takes effect on 12 November.
Anthropic published its 2026 Usage Policy update on 8 October, and it takes effect on 12 November. The company says most changes clarify existing rules rather than introduce new ones, drawing on misuse patterns from its September threat intelligence report and a year in which Claude took on longer, more independent work.
A new section, “Do Not Engage in Deceptive Campaigns or Artificial Activity,” consolidates rules against hiding who is behind a message or amplifying content through fake accounts, and against building infrastructure for influence operations. Claude is now explicitly barred from tracking people without consent, and from recommending who police should investigate, arrest or charge, though fraud monitoring, content moderation, journalism and legal research remain permitted. Where Claude connects to hardware capable of causing physical injury, operators must now keep it observed and have a safe-state failsafe to stop it. The weapons section now explicitly covers software and components that make weapons work, including arming autonomous vehicles. A new rule also bans sustained, needless cruelty toward the models, carving out frustration, testing and dark fiction.
Why it matters: if you’re building agents on Claude that touch physical hardware, scrape or profile people, or operate autonomously for long stretches, these are the specific boundaries you now need to design around before 12 November, not abstract “be safe” language.
The caveat: Anthropic also loosened one rule, dropping its blanket ban on personalised vote targeting so long as it isn’t deceptive or privacy-violating.