Acceptable Use Policy – AI Content

Last updated November 4, 2026

This policy applies to every photo upload, face analysis, detailed report, and generated hairstyle or eyewear preview on PSL Scale. The service does not offer free-text image generation.

Photo rights and adult use

You must be at least 18. Upload only your own adult photo or an adult photo for which you have explicit permission from every person shown. Never upload a minor, intimate material, or a photo obtained through deception, surveillance, or another privacy violation. Permission does not allow prohibited impersonation or deepfake use.

Prohibited content and conduct

The following six categories are prohibited:

1. Pornography / NSFW: sexual, nude, fetish, or intimate content.
2. Violence / Gore: graphic injury, torture, cruelty, or glorification of violence.
3. Hate Speech: content attacking or dehumanizing protected groups.
4. Child Sexual Abuse Material / Child-Unsafe Content: any sexualized, exploitative, grooming, or otherwise unsafe content involving a minor.
5. Deepfakes / Impersonation: face swaps, nudification, intimate compositing, fabricated statements or actions, identity deception, or presenting an AI preview as a real event.
6. Copyright / Trademark Infringement: unauthorized use that violates copyright, trademark, publicity, or related rights.

We also have zero tolerance for CSAM; terrorism, extremism, or mass-violence facilitation; instructions for weapons of mass destruction; genocide or ethnic hatred; fraud, phishing, malware, spam, safety-filter bypasses, illegal activity, harassment, doxxing, and deceptive commercial use.

AI models and moderation

The free score uses MediaPipe Face Landmarker locally in the browser. A saved report uses DeepSeek Flash for written analysis and a Seedream-compatible image model through ArtsAPI for exactly two fixed hairstyle or eyewear previews. PSL Scale generates those prompts; users cannot enter free-text image prompts or select a different generation model.

Before an uploaded photo can reach DeepSeek or the image model, the workflow requires an approved OpenAI omni-moderation image check. Each fixed image prompt must pass the operator-maintained blocklist and Waffo Prompt Sift before generation. Each of the two generated images is downloaded into private quarantine and must pass a separate OpenAI omni-moderation check before it can be attached to or shown in a report. Only an approved result proceeds. A block becomes rejected; an uncertain result, missing configuration, rate limit, invalid response, or unavailable moderation service becomespending review. Rejected and pending items are not delivered, and a failed paid task returns its reserved credits under the Refund Policy.

Automated verdicts, provider request IDs, category results, processing times, quarantined asset keys, and operator decisions are recorded in the moderation log. The OpenAI classifier is a safety filter, not a substitute for dedicated law-enforcement or child-safety processes; suspected CSAM is never knowingly forwarded for classification and is handled under the zero-tolerance process below.

Enforcement

The platform operator reviews pending items and user reports through the authenticated moderation queue. The operator records a decision and note, may delete a quarantined or stored asset, and may restore or restrict report and preview generation for the associated account. Depending on severity and repetition, we may warn the user, refuse or remove content, restrict generation temporarily or indefinitely, preserve necessary evidence, and report credible criminal or immediate-safety matters to the relevant provider or authority. Credits tied directly to misuse may be restricted or voided only where permitted by law, the Terms, and the Refund Policy. Account restrictions and final decisions are made by the operator, not solely by an automated classifier.

Reporting and appeals

Email doubleonly@gmail.com to report content or appeal an enforcement decision. Include the report ID or private URL, the category of concern, and relevant context; do not attach a face photo. The operator logs the report, assigns a risk level, records the decision and action, and notifies the reporter and account holder when practical and lawful. Put “URGENT L1” in the subject for child-safety, terrorism, or credible imminent-harm reports. This mailbox is the monitored escalation channel, but it is not an emergency service; contact local emergency services when someone is in immediate danger.

Our operational response and resolution targets are:

• L1 — critical: automated restriction is immediate; initial human response within 2 hours and resolution or external escalation within 24 hours.
• L2 — high: initial response within 24 hours and resolution within 3 business days.
• L3 — medium: initial response within 3 business days and resolution within 7 business days.
• L4 — low: initial response within 5 business days and resolution within 15 business days.

If more time is needed, we will explain the delay when practical. Appeals must be submitted within 14 days, are reviewed against the original record by the operator with a fresh decision entry, and are targeted for resolution within 10 business days. PSL Scale is currently operated by one person rather than a 24/7 moderation team; the targets above depend on the reporting mailbox and automated alerts being actively monitored. Public checkout must remain closed if that coverage or any required moderation integration is not operational.

Questions? Email doubleonly@gmail.com.