In plain English
- What happened
- Anthropic published a new version of its Usage Policy on October 8, 2026, set to take effect on November 12, 2026. The update rewrites the rules covering high-risk recommendations in health, legal and financial matters, consolidates rules on deceptive marketing and influence campaigns into one section, and clarifies existing prohibitions on weapons and surveillance uses.
- Why it matters to your business
- Any business using Claude to generate recommendations that touch a customer's health, legal rights, finances or livelihood now has a clearer statement of when a qualified person must review the output and when the customer must be told AI was involved. Businesses running social media or marketing campaigns should also note the new rule against using fake accounts or coordinated posting to amplify content, which now applies to commercial activity as well as political.
- Do this
- Open Anthropic's Usage Policy and Supported Regions pages and check the High-risk Use Case Requirements and Deceptive Campaigns sections against anything Claude currently touches in customer-facing health, finance, legal or marketing work, before November 12, 2026.
Anthropic revises its Usage Policy annually, and the 2026 version is built around how Claude is now being used for longer, more independent work. The company says most of the changes clarify existing rules rather than add new restrictions, drawing on patterns it documented in its own threat intelligence reporting over the past year.
What changes for health, finance and legal uses
The policy has long required a qualified human in the loop for Claude outputs that affect someone's health, legal rights, finances, livelihood or access to essential services, along with telling the affected person that AI was used. Those requirements have not changed, but Anthropic says users kept asking which use cases they actually cover, so the section has been rewritten to spell that out directly, listing which kinds of recommendations are included and which are not.
A new addition covers Claude connected to hardware that takes autonomous physical actions and could cause injury. Under the updated policy, a qualified operator must be able to watch the equipment and stop it if needed, and the equipment itself must be able to hold a safe state if Claude is disconnected.
Deceptive marketing and elections
Rules that were previously spread across Anthropic's elections, fraud, privacy and disinformation sections have been pulled into one new section called Do Not Engage in Deceptive Campaigns or Artificial Activity. It covers hiding who is behind a message, amplifying content through fake accounts, and building the infrastructure for influence operations, and it applies to commercial activity as well as political activity.
The elections section has been renamed Do Not Undermine Democratic Processes and narrowed to focus on deceiving voters or disrupting elections. Anthropic also removed a blanket ban on personalized voter and campaign targeting, which it says had been catching legitimate work such as nonprofits translating voter information or election officials sending ballot notices.
Weapons and surveillance language
The weapons section now states explicitly that the prohibition covers software and components that make weapons work, including arming drones or other autonomous vehicles. The surveillance section has also been rewritten to state more precisely that tracking people without consent is prohibited, whether in real time or through analysis of previously collected data, while permitted uses still include fraud monitoring, content moderation, journalism and legal research.
The updated policy takes effect on November 12, 2026.
Reality check
Anthropic describes the weapons, surveillance and high-risk sections as clarifications of enforcement it is already applying, not a change in practice. For most small businesses using Claude for ordinary customer service, content or admin work, this update changes the wording of the policy more than it changes what is allowed.
- 1
Check if Claude touches high-risk recommendations
Look at whether any workflow uses Claude to suggest something that affects a customer's health, finances, legal standing or job, since the current policy's human-in-the-loop and disclosure requirements apply there.
- 2
Review marketing and social media use
If Claude is used to generate or schedule social posts, confirm nothing in that workflow involves fake accounts, coordinated inauthentic posting, or obscuring who is behind a message.
- 3
Note the November 12 date
Put a reminder in the calendar to re-read the current Usage Policy and Supported Regions Policy before the update takes effect.
Sources
