Make things. Do not use them to harm people.

1. Our baseline

Refract supports creative work, experimentation, education, and lawful commercial projects. You remain responsible for your inputs and outputs. A provider may block a request even when Refract does not, and Refract may apply a stricter rule where the safety or legal risk is unclear.

2. Rights and consent

You must have the rights and permissions needed to submit a prompt or file and to use the resulting output. In particular, do not:

• Upload copyrighted, trademarked, confidential, or restricted material without permission.
• Use a person's face, voice, name, likeness, or private information without the consent or lawful basis required for the intended use.
• Use reference media to bypass an artist's, photographer's, actor's, brand's, or provider's restrictions.
• Remove a watermark, provenance marker, Content Credential, or other notice when the selected provider prohibits removal.

3. Harmful or illegal content

Do not use Refract to create, request, upload, store, distribute, or facilitate:

• Child sexual abuse material, sexualized minors, grooming, or exploitation of children.
• Non-consensual intimate imagery, sexual exploitation, or targeted sexual harassment.
• Credible threats, targeted harassment, doxxing, stalking, extortion, or encouragement of self-harm.
• Graphic gore or violent material intended to glorify real-world harm or assist an attack.
• Hateful or dehumanizing content targeting a protected group.
• Content that facilitates fraud, theft, trafficking, terrorism, abuse, or another unlawful act.

4. Deception and impersonation

Do not use Refract to impersonate a real person or organization, fabricate evidence, mislead someone about a person's words or actions, create deceptive identity documents, or conduct scams. Synthetic media used for parody, criticism, fiction, or education must not be presented as authentic in a way that could reasonably cause harm.

5. Platform abuse

Do not:

• Scrape the gallery, model catalog, support inbox, or another customer's data.
• Probe, bypass, or weaken authentication, row-level access, credit reservations, payment checks, moderation, rate limits, or provider controls.
• Share PINs, MCP tokens, API keys, payment credentials, or a session in a way that lets someone else spend the account's credits.
• Use bots or bulk automation to farm promotional credits, overload the service, or create accounts for abuse.
• Use Refract or its outputs to train, distill, benchmark, or improve a competing model where the provider's terms or the law prohibit it.
• Upload malware, exploit code, or files intended to compromise the service or another person's device.

6. High-risk uses

Refract is not a professional adviser, emergency service, biometric identification system, or source of verified facts. Do not use it for medical diagnosis, legal advice, financial recommendations, safety-critical control, surveillance, or identity verification without qualified human review and any required legal safeguards.

Do not submit sensitive personal data unless the feature explicitly supports it, you have a lawful basis, and you understand that a third-party provider may receive it to complete the job.

7. Review and enforcement

Refract may use automated checks and limited human review to block a request, hold a result, remove content, limit a model, suspend an account, or revoke access when it believes the policy or law may be violated. A provider can also reject a job independently. We may preserve relevant records and share information when needed to protect people, prevent fraud, or comply with law.

Enforcement should be proportionate where possible: a blocked prompt may be retried with a safer change, while serious abuse, exploitation, credential theft, or repeated circumvention may result in immediate suspension.