Anthropic is tightening its rules around deceptive content designed to influence what search engines and AI systems say.
An updated usage policy, set to take effect Nov. 12, prohibits using Claude to create or distribute content that misrepresents its origin, authorship or independence in order to influence the sources used by search and AI systems.
The new rules specifically prohibit networks of websites that appear independent but are actually coordinated to reinforce the same claims.
Anthropic also bans the use of Claude to create fake personas, reviews, news outlets or accounts designed to hide their shared origin.
The restrictions sit within a broader section covering deceptive campaigns and artificial activity, applying to both commercial and political use cases.
Anthropic says similar activities were already restricted across different parts of its policies, but the updated rules bring them together under a clearer framework.
The policy applies across Anthropic products, including Claude applications and API-based services. Violations could result in suspended or terminated access.
As AI systems increasingly rely on web content to generate answers, influencing the underlying source environment is becoming a new form of information manipulation.
Anthropic’s policy signals that AI governance is expanding beyond what models generate to how they may be used to shape the information those models later retrieve and trust.
