Anthropic, the developer of the Claude neural network, is significantly tightening its platform usage rules, effective November 12. It is now explicitly forbidden to use Claude to “seed” search engine and AI system sources with misleading content. This applies to materials that misrepresent their origin, authorship, or independence. Imagine a network of websites that pretend to be independent but promote the same claims – this is what Anthropic will fight against, striving to ensure the purity of the information space.
Anthropic tightens rules: fighting fake sources for AI answers
New section “Prohibition of misleading campaigns”
The key change in Anthropic’s policy is the introduction of a new section titled “Do not engage in deceptive campaigns or artificial activity.” This section clearly states the unacceptability of manipulating sources from which search engines or AI systems draw answers. For example, creating networks of websites that pretend to be independent sources but confirm the same claims is now strictly prohibited. This is critical for maintaining trust in information generated by neural networks.
- Prohibition of using fake personas, publications, and reviews.
- Dissemination of content through “fake or supposedly independent publications, websites, or accounts” with the aim of concealing a common source.
- Prohibition of creating tools for deceptive campaigns and selling such services.
Previously, similar rules were scattered across various sections of the policy concerning elections, fraud, privacy, and disinformation. Now everything is collected in one place, which significantly simplifies the understanding and application of the new norms.
Lessons from September: real cases of deception
Anthropic is not just introducing new rules, but also reinforcing them with real examples from its September threat analysis report. The company identified about 70 websites masquerading as independent local news outlets. All of them were linked to one advertising agency from France.
“The operation involved using Claude to create articles in a uniform format, each containing three to four internal links, for automatic publication.”
These articles, of which there were at least 8913 in approximately 20 languages, were specifically designed to increase the authority of the sites in search engines. Although most of them did not receive significant audience response, Anthropic promptly responded by blocking the Claude account and the organization associated with this activity.
Automated Publishing: No Longer “High Risk”
An interesting change has affected automated content publishing. Previously, “automated generation and publication of content for external consumption” was considered a high-risk scenario. This point has now been removed from the list. This does not mean complete permissiveness, but indicates that Anthropic sees potential in mass AI video generation and other content formats, provided they do not violate other rules, especially regarding misinformation.
The list of high-risk areas requiring review and disclosure now includes 11 items, such as legal, medical, and financial advice, as well as decisions about credit, housing, and employment. The requirements for reviewing recommendations in these areas remain unchanged: a qualified professional must review the findings, and users must be notified that the content was created with AI involvement. This opens new horizons for AI video for ads and AI UGC for crypto projects, where AI-powered subscription production could become the standard.
Google and Anthropic: A United Front Against Spam
Google is also actively fighting the manipulation of generative AI responses in search, calling it spam. An updated document from May 15 clearly stated that these policies apply to generative AI responses. While Google applies its rules to websites, which can lead to their ranking being lowered or removal from search results, Anthropic acts at the user level for Claude, restricting or revoking access. This demonstrates a unified approach by tech giants to ensure the integrity of the information space.
Frequently Asked Questions
What has changed in Claude’s usage policy?
The main change is the prohibition of using Claude to create and disseminate misleading content that manipulates information sources for search engines and AI. A new section on “Prohibition of Misleading Campaigns or Artificial Activity” has been added.
Why did Anthropic introduce these changes?
The changes are aimed at combating misinformation and manipulation that can affect search engine and AI responses. This is a response to identified cases where Claude was used to create networks of fake news websites.
How will this affect AI content creation?
This does not prohibit AI content creation as such, but establishes strict frameworks for its use. For example, mass video generation by neural networks or the creation of AI avatars with subtitles are still possible if the content does not mislead about its origin or authorship. It is important to be transparent and ethical.
What are the “high-risk” areas of Claude’s use?
High-risk areas include legal, medical, financial advice, as well as decisions affecting credit, housing, and employment. In these cases, mandatory verification by a qualified specialist and user notification that the content was created using AI are required.
Is automatic content publishing now prohibited?
No, automatic content publishing is no longer considered “high-risk.” This opens up opportunities for an AI pipeline of 500 videos per day or an AI video factory for e-commerce, provided all other rules are followed, especially those concerning misinformation and manipulation.
Conclusion
Anthropic’s updated policy is an important step towards more ethical and responsible use of generative AI. This is not only about protecting users from misinformation but also about creating a more transparent environment for technological development. For businesses that use AI video for advertising or neural networks for UGC advertising, this means being even more attentive to the origin and presentation of content. We see that the effectiveness of AI versus real UGC continues to grow, but with it, so does responsibility. It’s time to adapt and build a future where technology serves for good, not for deception. If you are looking for where to order AI UGC for crypto or other projects, make sure your partner adheres to high standards of ethics and transparency.