Anthropic is tightening the rules around Claude access. Sharing your consumer subscription as a business model is now off-limits.
Anthropic has updated its Acceptable Use Policy (AUP), covering Claude[.]ai, Claude Code, API access, cloud platforms, and products integrating Claude.
11 high-risk domains are covered, including law, healthcare, finance, credit, and hiring.
AI-generated recommendations in these areas require qualified human review and disclosure of AI involvement.
AI-controlled hardware capable of causing physical injury must have human oversight and an immediate stop mechanism.
Unauthorized use of Claude outputs to train or refine models is prohibited.
Reselling or transferring Claude access through consumer subscriptions is prohibited.
Law enforcement decisions and non-consensual surveillance using Claude are restricted.
Cybersecurity and life-science users can apply for adjustments to certain safety limits.
Market angle: This is about more than subscription sharing. Anthropic is drawing clearer boundaries around commercial access, model-output exploitation, and high-risk AI deployment. It could affect third-party services that rely on consumer Claude subscriptions to provide access to customers.
Your Claude subscription is not a cloud business license.
Will tighter access rules protect the AI ecosystem, or create more friction for developers building on top of Claude?
Anthropic has updated its Acceptable Use Policy (AUP), covering Claude[.]ai, Claude Code, API access, cloud platforms, and products integrating Claude.
11 high-risk domains are covered, including law, healthcare, finance, credit, and hiring.
AI-generated recommendations in these areas require qualified human review and disclosure of AI involvement.
AI-controlled hardware capable of causing physical injury must have human oversight and an immediate stop mechanism.
Unauthorized use of Claude outputs to train or refine models is prohibited.
Reselling or transferring Claude access through consumer subscriptions is prohibited.
Law enforcement decisions and non-consensual surveillance using Claude are restricted.
Cybersecurity and life-science users can apply for adjustments to certain safety limits.
Market angle: This is about more than subscription sharing. Anthropic is drawing clearer boundaries around commercial access, model-output exploitation, and high-risk AI deployment. It could affect third-party services that rely on consumer Claude subscriptions to provide access to customers.
Your Claude subscription is not a cloud business license.
Will tighter access rules protect the AI ecosystem, or create more friction for developers building on top of Claude?
