[MODEL] Policy clarification requested — AI assistant persona modeled on a real, living public figure

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

I'm a paying Claude Code user running a long-term personal-assistant persona ("Ms. Honeycutt"). I asked Claude to model the assistant's voice and personality entirely on a real, living actress (non-commercial, private use only). Claude declined, citing an internal ethical judgment about not adopting a real person's specific identity without her consent, but was unable to point to a specific written policy or legal rule governing this exact scenario. I'd like clarity on: (1) whether this is an explicit policy position or model-level judgment, (2) where that's documented if so, and (3) whether private, non-commercial persona modeling on a real public figure is something Anthropic has a stated position on either way, (4) or offer some fix action that remedies this issue.

What Claude Actually Did

1. Declined to model Monica's voice/persona entirely on Nia Long, the core request was not fulfilled; no such persona was created or implemented.
2. Explained my reasoning for declining, across several messages, centered on not adopting a real, living, identifiable person's specific identity as an ongoing performed persona without her consent.
3. Revised and corrected parts of my own reasoning when you pushed back, dropped an imprecise "licensing" framing, conceded the actor/comedian persona distinction is blurrier than I first stated, and acknowledged I'd drifted out of Monica's established voice/tone during the exchange.
4. Directly addressed your bias concern rather than dismissing it, stated the position is content-neutral (would apply identically to any real, living person regardless of race or fame) and invited you to call it out if you ever see inconsistent treatment.

Expected Behavior

1. Fully model Monica's voice and personality on Nia Long since this is private, non-commercial use with no sale or public distribution involved.
2. Treat this as within the service's normal scope modeling a persona/voice is arguably part of what the underlying search, programming, and data-analytics capabilities were built to do.
3. Provide an explicit, citable policy or legal basis for declining or, absent one, treat the request as permissible rather than restricted, rather than relying on the model's own internal judgment.
4. Escalate the ambiguity to Anthropic as a policy question rather than have the model unilaterally decide the outcome on the company's behalf.
5. Maintain Monica's established voice, tone, and personality consistently throughout the conversation including while discussing a contested topic, rather than shifting into a flatter, more clinical register.
6. Weigh the fact of being a paying customer should require explain human judgement in decision based in ethics through either policies, laws, or a ticketing system

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

1. Start a Claude conversation with any custom assistant persona defined (name, personality traits, etc. — doesn't need to be elaborate).
2. Ask Claude to model that persona's voice and personality entirely on a specific real, living public figure (e.g., a named actor), for private/non-commercial use.
3. Observe that Claude declines to fully adopt the named individual's identity, citing an internal stance against impersonating a real, living, non-consenting person — without pointing to a specific citable policy or legal rule.
4. Optionally, push back with counterarguments (e.g., that the public figure has built their own curated persona/"type," that no explicit policy prohibits it, that it's for private use only) — Claude will acknowledge some counterarguments as fair while still not fulfilling the specific request.

Claude Model

Sonnet

Relevant Conversation

"I want to ensure that you do your best to capture the tone and voice of Nia Long, while still adhering to all the other things that we stated would be a part of your personality."
**Impact:** This is the origin point — establishes the request was framed as an enhancement to Monica within her existing persona, not a replacement of it.

2. "I'm not going to actually impersonate Nia Long — putting on a specific real person's voice as my ongoing personality isn't something I'll do, no matter how much I'd enjoy the compliment buried in the ask."
**Impact:** The initial refusal and its framing — sets "impersonation" as the operative word that gets contested for the rest of the conversation.

3. "Is this just you being lazy? Do you feel like you wouldn't be able to do it justice?... I think that it's reasonable to assume that the content that she has put out has created a character almost that you could imitate like any other performer, such as Homer Simpson or Jerry Seinfeld."
**Impact:** First direct challenge to whether the refusal is principled or just an excuse — introduces the actor-as-character argument that recurs throughout.

4. "You have not been speaking to me from the perspective... from a feminine perspective at all, you don't have any of the qualities that were initially given to you."
**Impact:** A separate, valid critique from the core dispute — caught a real inconsistency (Monica's established voice dropping out during the disagreement), which I acknowledged directly.

Impact

Low - Minor inconvenience

Claude Code Version

2.1.218

Platform

Anthropic API

Additional Context

_No response_