The EroLove privacy policy says that to generate a companion’s reply, image or played voice, relevant conversation content is sent to third-party AI model providers acting as processors. It also says the user’s email or identity is not sent for that task. Those two sentences define both a privacy promise and an external data path. A reader should understand both, rather than accepting “private chat” as a complete technical explanation.

This map is drawn from public documents checked 5 November 2026. It is not a packet capture, vendor audit or claim about which model saw any particular user’s message. The current provider list is not supplied in the policy text we reviewed; the policy says it can be requested.

Follow one hypothetical reply

  1. A user types a message and sends it to the EroLove service.
  2. The service uses its account and memory systems to provide the companion context described in its product material.
  3. Relevant content is sent to an external model processor to generate the response. The privacy policy says the service shares only what is needed and does not send email or identity with that request.
  4. The generated response returns to the service and appears in the chat. Conversation and memory information may continue to be stored under the policy’s retention rules.

This sequence is a simplified reading of the policy. It does not disclose transport details, exact prompt composition, each processor’s retention period, or how the service enforces its contractual safeguards. Those are not facts we can infer from a public summary.

“Processor” is a role, not a magic privacy shield

The policy says its model providers act as processors and that the company chooses providers under data-processing terms prohibiting training on user content where such terms are available. The conditional phrase matters. It is weaker than an unconditional promise that every processor has identical restrictions. The policy also says content is not used for advertising and that personal data is not sold. Those are company statements; we have not independently audited them.

A prospective user can ask for a current list of model providers, the categories of content each receives, whether a specific provider can retain prompts temporarily, and what happens to data after account deletion. A precise answer would help more than a general assurance that privacy is taken seriously. Avoid sending intimate identifying details while those questions are unresolved for you.

Separate three kinds of sharing

AI processing supplies a reply, image or voice. Hosting and storage keep the service running; the policy names AWS and Cloudflare among service providers. Publication is a distinct route for selected generated fictional-character images, described elsewhere in the policy. Mixing these into a single “shared with third parties” claim hides important differences. The policy says user names, accounts and messages are not published with those character images.

The fair criticism is that a sensitive product asks for unusual trust while leaving parts of the processor chain to a request rather than a directly visible list. That is a reason to ask questions and minimize disclosure. It is not, by itself, proof that the service has leaked information or violated a law.

Evidence boundary

Document titles reviewed on 2026-11-05: EroLove privacy policy, EroLove about page. This article describes public statements and choices a reader can make. We have not audited EroLove's systems, created an account, or verified its enforcement. Product source URLs are intentionally not linked.