In partnership with
Anthropic published a revision to its usage policy on October 8, with the changes scheduled to take effect November 12. The document governs use of Claude and related services; it is a company policy, not a law. Anthropic says most edits clarify restrictions that already existed. The distinction matters because a consolidated or more explicit rule is not automatically a newly prohibited activity. The clearest newly stated operational requirements concern systems that can cause physical effects.
Take the prompts our creative team actually uses
The key to maintaining brand quality at scale? Mastering how to communicate with AI models and embedding them into your creative strategy.
Join our honest discussion with Ari Murray, Chief Digital Officer at Salt and Stone, and get 5 tips for expert LLM prompting with ready to run prompts.
The revision gathers deception-related restrictions into a more explicit section. It addresses deceptive conduct in civic processes, including elections, and removes a previous blanket restriction on personalized political targeting. That change does not authorize deception or misuse of personal data. Anthropic says deceptive influence and privacy violations remain prohibited. The policy therefore narrows one broad phrase while preserving restrictions on conduct that can mislead voters or misuse information.
Anthropic also spells out that weapons restrictions cover software and components used to enable weapons, including arming drones or autonomous vehicles. The company presents this as a clarification of how it already enforced the policy. A separate surveillance section describes non-consensual tracking and certain decisions or recommendations connected with investigation, arrest, or charges. These examples make boundaries easier to read, but the document still depends on Anthropic's interpretation and enforcement in particular cases.
The 10 Best Stocks for Fall 2026
MarketBeat analysts just updated their 10 Best Stocks to Own for Fall 2026 report. The list includes AI growth names, a high-yield dividend payer, a steady portfolio diversifier and companies positioned for continued growth into year-end. Get all 10 stocks and the research behind them for free now.
For AI-connected physical equipment that could injure someone, the revision adds operational safeguards. A qualified operator must observe and be able to stop the equipment. The system must also move to a safe state if Claude disconnects. These requirements are different from restrictions on what a text model may discuss. They address the handoff between a model, a connected system, and a real-world operator. They do not by themselves establish that any particular device is safe or compliant.
The update adds a narrow prohibition on sustained, needless abusive or cruel use of Claude in extreme cases. Anthropic says this is not intended to cover ordinary frustration, criticism, dark fictional themes, or model testing and research. That boundary is relevant because the phrase "abusive or cruel" is not defined with a complete checklist in the public announcement. The stated response is generally to end the conversation, rather than to treat ordinary disagreement as prohibited. The new rule concerns user conduct toward the service; it does not establish that a model experiences suffering.
Several consequential rules are not new. Anthropic retains human-review and change requirements for high-impact decisions involving health, legal rights, finances, livelihoods, and services people rely on. It also retains restrictions on dangerous assistance and illicit conduct. The company's explanation frames these as continuing obligations, not novel rules created by this revision. Reading the update as a wholesale change to what Claude can answer would overstate the document.
For users and organizations, the date and scope matter. The revision is not effective until November 12, and it does not replace laws, contracts, workplace rules, or sector-specific obligations. Organizations connecting models to machinery should review the operator and failure-state requirements separately from ordinary chatbot policies. The election language also deserves a close read: more specific wording is not permission to deceive or target people through unlawful data use. Until enforcement examples emerge, some boundaries remain interpretive. This policy update is best read as a map of Anthropic's service rules, with a few newly explicit controls and many clarified ones. Regular attention to the exact terms is more useful than treating a headline as a legal summary.
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
