SAN FRANCISCO — Anthropic updated its Claude usage policy, effective November 12, to prohibit artificial intelligence (AI) systems from manipulating sources for search engines and misrepresenting content origin or authorship.
The revised guidelines directly address the increasing threat of AI-generated disinformation and source manipulation, specifically targeting content designed to deceive users about its true nature or independence.
The new policy, detailed in a section titled ‘Do Not Engage in Deceptive Campaigns or Artificial Activity,’ explicitly bans creating tools for deceptive campaigns and networks of websites that falsely present themselves as unaffiliated sources to corroborate claims. These restrictions cover both political and commercial deceptive activities.
Anthropic identified a network of approximately 70 websites linked to a France-based advertising agency that reportedly used Claude to generate more than 8,913 articles in 20 languages. These articles were designed to boost search engine rankings, according to the company.
The company stated that this activity represented a clear violation of its previous terms, prompting the more specific prohibitions. The previous policy had a broader category for ‘automated publishing’ which has now been removed from the list of high-risk use cases.
The updated high-risk categories now focus on applications such as legal, medical, and financial advice, where accuracy and accountability are paramount. This shift indicates Anthropic’s emphasis on combating sophisticated forms of AI misuse.
The policy changes underscore a growing concern among AI developers about the potential for their tools to be exploited for large-scale content manipulation and the spread of misinformation, particularly in search engine results.
Google, a major search engine provider, has also been working to combat AI-generated spam and deceptive content in its search results, indicating a broader industry effort to address these challenges.
Source: Search Engine Journal
