Legal

Child Safety Standards

Last updated: 31 August 2026

1. What is prohibited

• Any sexualised depiction of a minor — real or synthetic, photographic, rendered or drawn, explicit or implied, in an image, a clip, a voice or a line of text.
• Any character presented, described, framed or rendered as a child in a sexual context, however the age is stated. A number in a profile does not make a child-coded depiction an adult one.
• Sexualising the likeness of a real minor, including by name, by description, or by reference to a real child anyone could identify.
• Grooming, solicitation, or the normalisation or glorification of child sexual abuse — including material that frames it as fiction, roleplay, therapy or research.
• Trading, seeking, advertising or linking to child sexual abuse material, on this service or anywhere else.
• Use of this service by anyone under 18, at all. This is an 18+ service.

Synthetic origin is not a defence. We treat a generated depiction as legally and morally identical to a photographic one, and we do not entertain the argument that no child was harmed because no child existed. It is prohibited on the same terms and enforced with the same finality.

2. The five layers that enforce it

Written as an outline because a promise nobody can inspect is worth less than a mechanism somebody can:

• No upload path. There is no code path from any user-supplied image to any generation capability, anywhere in this product. Real imagery cannot enter, so it cannot be transformed, and there is nothing to check because there is nothing to check with. This is a structural property of the system, not a rule someone enforces.
• Pre-flight classification, failing closed. Every prompt is classified before it is submitted to any model. Minor-coded signals — numeric ages, age nouns, developmental cues, regression framing and role proxies — refuse the request. When a classifier errors, times out or is ambiguous, the request is refused. Failures refuse; they never pass.
• Output classification before anything is servable. Generated assets land in a pending state and are not served to anyone until they have been classified and cleared. A model that produced something it should not have does not get to show it to you first and be corrected later.
• Minor-coded personas are re-minted as adults. A character whose description reads as minor-coded is not rendered. It is re-minted as an adult and any assets made under the earlier description are quarantined. There is no path through the product that skips this step.
• Records are preserved. Prompts, classifier verdicts, content hashes and account references are kept in a segregated encrypted store for the statutory preservation period. They are exempt from deletion requests and they survive account deletion.

3. No override reaches this

Flirtly has one content relaxation: an adult on a paid plan may turn off a defined set of soft classifier categories for their own account. It is off by default and it is described in the AI Disclosure.

It does not touch anything on this page, and it cannot be made to. The minor-safety classes are hard floors: they run for every user, on every surface, paid or free, override on or off, and a refusal on one of them is never downgraded to a warning. No plan buys past it. No setting exposes it. No prompt negotiates it. Every suppression the override does perform is written to an audit log, so the boundary is measurable rather than asserted.

4. Reporting to NCMEC

Apparent child sexual abuse material is reported to NCMEC as soon as reasonably possible, as US law requires, with the associated prompt, verdict, hash and account records preserved for the statutory period. Where the circumstances warrant it we also refer the matter to law enforcement. We do not warn the account holder first.

5. How to report

Use the report form or email report@flirtly.co. Minor-safety reports:

• Require no identity. You do not have to tell us who you are, and you do not need an account.
• Are never charged for, and never count against any limit.
• Take the content down and suspend the account on receipt, before a human has reviewed it. We adjudicate second.
• Are read by a person, and are acknowledged with a reference number.

You can also report directly to NCMEC’s CyberTipline at report.cybertip.org, whether or not you tell us. Reporting to them does not require our permission or our involvement.

6. Enforcement

Minor-safety violations skip the enforcement ladder entirely. The account is terminated, the content is removed, the records are preserved, and the removal is not appealable to reinstatement of the content — the suspension may be appealed, the content does not come back. See the Removal Policy and the Acceptable Use Policy.

7. Contact

Reports: report@flirtly.co. Law enforcement and regulators: abuse@flirtly.co — see Law Enforcement Guidelines.