Anthropic usage policy update: what changes for Claude apps
Anthropic's usage policy update takes effect November 12. New rules on fake accounts, tracking, and physical agents. Here's what to check in your Claude app.
Anthropic published a usage policy update today, effective November 12. The headlines are about a new rule against abusing Claude itself. That rule will not touch your business. The other changes might, if you run Claude in a marketing tool, a monitoring workflow, or anything that moves physical equipment. You have five weeks to check.
What actually happened
Per Anthropic's announcement and TechCrunch's report:
- New "Deceptive Campaigns or Artificial Activity" section. It pulls rules from the old election, fraud, privacy, and disinformation sections into one place. It covers commercial activity as well as political: fake accounts, fabricated outlets, and tools for influence operations.
- Elections section narrowed. Now "Do Not Undermine Democratic Processes," focused on deceiving voters and disrupting elections. The blanket ban on personalized campaign targeting is gone.
- Surveillance spelled out. No tracking people without consent, live or from data you already collected. Consented uses, like fraud monitoring, stay allowed.
- High-risk use cases clarified, not changed. A qualified human must still be able to review and change Claude's recommendations, and affected people must be told AI was used.
- New rules for physical actions. If Claude drives hardware that could injure someone, a qualified operator must be able to watch and stop it, and the equipment must hold a safe state if Claude disconnects.
- Abuse of the model. Banned only in extreme, sustained cases. Normal frustration and testing are explicitly excluded.
Why it matters for your business
Fake reviews and sock-puppet accounts are now named commercial violations. If any tool you build or buy uses Claude to post reviews, comments, or social replies under made-up identities, stop. This was risky before. Now it's written down.
Customer tracking needs consent. Feeding order history or location data into Claude to profile individuals without consent is now an explicit line. Fraud checks are fine. Covert profiling is not.
High-risk decisions still need a human with the power to override. Lending, hiring, housing, insurance: if Claude recommends, a person decides, and the customer gets told.
Warehouse and shop-floor agents need a kill switch. If your agent runs a printer, a conveyor, or a robot arm, build the stop button and the safe state before November 12.
Key takeaways
- Anthropic's updated usage policy takes effect November 12
- Fake accounts and artificial activity are now banned for commercial use too, not only politics
- Tracking people without consent is prohibited; consented fraud monitoring is allowed
- High-risk recommendations still need human review and disclosure to the person affected
- Claude-driven hardware needs an operator who can stop it and a safe state on disconnect
Not sure your Claude workflow passes? We build agents with approval steps, audit logs, and stop controls from day one, so a policy change is a quick review instead of a rebuild. Send us what your agent does and we'll tell you what to change, or see how we build.
Sources: Anthropic, TechCrunch.
- #anthropic
- #claude
- #usage-policy
- #ai-compliance
- #ai-agents
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
nCino + Parlay: AI SBA loan intake comes to 1,400 lenders
nCino will resell Parlay's AI SBA loan intake to banks. If you plan to borrow, an AI now screens your file first. Here's how to make it credit-ready.
Read itGemini coworker agents get email and Drive: onboard them like hires
Google's Gemini at Work 2026 gives coworker agents their own email, calendar and Drive. Treat each one like a new hire: scoped access, a budget, an audit log.
Read it