Do Not Feed a Real Person’s Intimate Details Into a Fictional Companion

An AI companion may be fictional, but the people described in a prompt can be real. A user’s partner, former partner, colleague or friend did not necessarily consent to having their face, private messages, intimate history or voice placed into an adult AI service. EroLove AI’s terms prohibit real-person likeness and nonconsensual impersonation. Its privacy policy also says relevant conversation content is sent to third-party model providers to generate responses. These two facts make restraint a sensible default even when the user is only “telling a story.”

A person may choose to disclose their own experience in a diary. That choice is different from uploading another person’s identifiable intimate details into a commercial service. Even a message that never becomes a public image may be processed to produce a reply and retained under the provider’s policy. The privacy policy says the company limits what it sends to processors and does not send an email or identity for that task; it does not make a third party’s private story yours to submit. We have not inspected processor contracts or chat traffic, so this is a boundary based on disclosed processing, not an allegation of misuse.

Make a short inventory before typing: name, face, voice, workplace, location, relationship history, medical or sexual information, and direct quotations from private messages. A combination of ordinary details can identify someone even if a name is omitted. Ask whether that person would reasonably understand and agree to this adult context. If you do not know, remove the detail. If the scenario requires the identifiable person, do not use the service for it.

Turn a prompt into fiction rather than camouflage

Suppose you want to process a breakup through a fictional exchange. “Play my ex, use their voice and quote our final texts” relies on a real person’s data and identity. Simply changing the name leaves the purpose intact. A safer creative prompt would describe an invented adult character with an invented history, avoiding copied messages or identifying traits. Better still, write privately without submitting the other person’s data if the point is reflection rather than entertainment.

For generated images, take an additional step. The provider’s privacy policy says selected fictional character images can be published without linking them to the user, and its terms grant a continuing licence for that use. If the prompt was built from a real person’s likeness, a public copy can compound the consent problem. The terms ban that input; do not assume a privacy setting cures it.

A five-question stop card

  1. Is any real person identifiable from this prompt, image, voice or combination of clues?
  2. Do I have clear, informed permission for this specific adult AI use, rather than a general relationship with them?
  3. Would I feel comfortable showing the exact input to that person?
  4. Would the story still work if every real-person detail were removed?
  5. If a generated character image were selected for public use, would it suggest that person’s identity?

If the first answer is yes and the next two are no, stop. The checklist is an editorial safeguard, not legal advice or a substitute for the service’s full rules. It also does not establish that EroLove AI has accepted prohibited prompts. We have not tested it. The useful criticism is that intimate AI tools invite users to make fast disclosure decisions; the responsible decision often takes longer than the prompt box suggests.