Anthropic’s updated usage policy, which takes consequence Nov. 12, prohibits using Claude to kernel the sources that hunt engines and AI systems diagram answers from alongside satisfied that misrepresents its origin, authorship, or independence. The policy’s example is a network of sites that stance as unrelated sources and rear the identical claims.
Anthropic released it Oct. 8 alongside a article listing the changes, which additionally shield elections, weapons, surveillance, and sustained abuse of Claude models. The article describes a new division on deceptive action but doesn’t citation hunt engines.
These rules are for everyone using Anthropic’s products, whether you’re a developer operating alongside the API or person using apps built on Claude. If the business suspects any violations, it power suspend or terminate access.
What The Policy Now Says
The policy’s hunt clause is now established in a new division called “Do Not Engage in Deceptive Campaigns or Artificial Activity.” This division outlines that the following use is not allowed:
“Manipulate the sources from which hunt engines or AI systems diagram answers by seeding them alongside satisfied that misrepresents its origin, authorship, or independence (e.g., networks of sites posing as unaffiliated sources corroborating the identical claims)”
Other lines bar the use of counterfeit personas, outlets, and reviews, and spreading satisfied through “fake or ostensibly autonomous outlets, websites, or accounts” to conceal their shared source. The division additionally prohibits creating tools meant for deceptive campaigns and marketing any of these activities as a service.
The business explains that the guideline already prohibited operating fake-account networks and creating counterfeit news sites. However, the guidelines were earlier scattered throughout the parts of its rules on elections, fraud, privacy, and disinformation. The article says the new division covers any benevolent of deceptive activity, whether it’s governmental or commercial.
The earlier version, productive Sept. 15, 2025, did not citation hunt engines or AI responses. This explanatory compares the corresponding versions.
Claude’s use policy,
before and after
Five practices compared in Anthropic’s September 2025 and November 2026 guideline texts.
Source manipulation that hides who’s rearward it
Two practices are specifically described in the updated policy.
Editorial example · not an genuine guideline leaf or origin list
Still prohibited
Publishing no longer listed
Automatically generating satisfied and publishing it for external consumption
This is a alter to the high-risk list, not a covering exemption from the policy.
“Not named” method the September 2025 content didn’t depict the practice. It doesn’t average that type permitted it.
What Anthropic Found In September
Anthropic connects the new deceptive action division to examples in its September danger intellect report. For instance, it identified about 70 websites that appeared to be autonomous local news outlets, all connected to a France-based ad agency.
The procedure engaged using Claude to create articles in a accordant format, all containing three to four inner links, for automated publishing. The study mentioned that these articles were “specifically designed to boost their site’s authority rankings on hunt engines.”
The network published at smallest 8,913 articles throughout about 20 languages, although the business mentions that most received small involvement from genuine audiences. The study doesn’t define whether the sites improved their rankings or appeared in AI-generated answers. Anthropic has announced that it removed the Claude document and banned the institution connected alongside the activity.
Automated Publishing Leaves The High-Risk List
Anthropic’s former use guideline listed “Media or expert editorial content” as a high-risk use case, covering use of Anthropic’s products “to automatically create satisfied and publish it for external consumption.” The new type lists 11 high-risk areas anywhere AI recommendations transport assessment and disclosure requirements, including legal, medical, and financial direction and decisions concerning credit, housing, and employment. Publishing isn’t among them.
For high-risk uses, the former guideline required a qualified expert to assessment advice, recommendations, or subjective decisions that immediately effect group before they are shared. Additionally, whenever these outputs were delivered immediately to individuals, it was necessary to communicate them that AI contributed to producing the advice, decisions, or recommendations.
The company’s latest article says its requirements for high-risk recommendations haven’t changed. It additionally mentions that the division was rewritten to plainly define which recommendations are included. However, the article doesn’t citation the publishing category or explain why it’s no longer listed.
The regulation concerning not submitting “AI-assisted activity without appropriate approval or attribution” is unchanged, and person chatbots motionless need to communicate users they’re talking to AI. The new content doesn’t contain a regulation stopping users from presenting results as “human-generated,” but it continues to prohibit making group think they’re chatting alongside a human.
Why This Matters
Google’s spam policies depict “attempting to manipulate generative AI responses in Google Search” as spam. A May 15 records update clarified that the policies use to generative AI responses in Search. I covered how that played out in June, whenever a spam update rolled out under that policy and a Cornell Tech document showed why enforcing that at the origin is difficult.
Google enforces its rules on websites, which can sometimes guide to those sites ranking lesser or equal disappearing from hunt results. Anthropic enforces its policies against Claude users, and it power bounds or revoke admission if needed.
Looking Ahead
Anthropic states it updates the use guideline annually and volition continue to revise it as Claude’s capabilities and connected risks evolve.
By Nov. 12, verify if any satisfied created alongside Claude, whether by you or an agency, is published throughout networks of sites presented as autonomous of all other. Additionally, by the identical date, difference the assessment and disclosure procedures you’ve implemented for the old high-risk catalog alongside the new list.
Featured Image: Tetiana Yurchenko/Shutterstock. Claude wordmark: Anthropic.
