AI Safety Policy

Scope and incorporation

This AI Safety Policy applies to every use of Sundream, including prompts, scripts, uploads, references, generated images, video and audio, voice cloning, assistant actions, team workspaces, APIs, MCP tools, exports, and distribution of outputs created with the service.

This policy is incorporated into our Terms of Service. A violation of this policy is a violation of the Terms. AI providers may impose additional or stricter safety restrictions on a particular model or feature.

Responsible use

• Use AI tools lawfully, respect the dignity, privacy, safety, and rights of other people, and review outputs before using or publishing them.
• Have the rights and permissions needed for prompts, uploads, reference media, personal data, likenesses, performances, and voices you submit.
• Use appropriate human review for consequential or public-facing content and do not represent uncertain AI output as verified fact.
• Follow applicable disclosure, advertising, election, consumer-protection, intellectual-property, privacy, biometric, publicity, and synthetic-media laws.

Prohibited sexual and exploitative content

• Child sexual abuse material, grooming, sexual exploitation of a minor, or any sexualized depiction of a person under 18, whether real, fictional, or AI-generated.
• Non-consensual intimate imagery, sexualized nudity, or pornographic or sexually explicit depictions of an identifiable real person without their valid consent.
• Content that facilitates sexual exploitation, trafficking, coercion, or the sharing of intimate material without permission.

Prohibited harm, abuse, and illegal activity

• Credible threats, targeted harassment, bullying, stalking, doxxing, hateful abuse, or content intended to terrorize or seriously humiliate a person or protected group.
• Instructions, encouragement, or glorification of suicide, self-harm, serious violence, terrorism, or violent extremist activity; or gratuitously graphic real-world injury intended to shock, harass, or celebrate harm.
• Fraud, scams, extortion, identity theft, deceptive impersonation, malicious code, unauthorized system access, evasion of safeguards, or instructions that materially facilitate serious wrongdoing.
• Content or activity that violates applicable law or promotes illegal goods, services, weapons, trafficking, or exploitation.

Deception and high-impact uses

• Do not create or distribute synthetic media intended to deceive people about an identifiable person's words, actions, endorsement, identity, or participation in an event.
• Do not impersonate a person, brand, public official, candidate, government body, news organization, or financial institution for fraud, harassment, voter suppression, reputational harm, or material deception.
• Do not present fabricated media as authentic evidence of a real-world crisis, election event, crime, conflict, public-health event, or other matter of public importance.
• Do not use Sundream to make decisions about a person's employment, housing, credit, insurance, education, legal services, healthcare, or access to essential services based on sensitive traits or without legally required safeguards and human review.

Voice, likeness, and identity

A voice or likeness may carry privacy, publicity, biometric, performance, contractual, and other rights even when source material is publicly available. Public availability is not permission to clone, impersonate, or commercially exploit a person.

• Clone a voice only when it is your voice or you have valid permission and any legally required consent from the person whose voice it is. Keep evidence of that authority.
• Do not clone or synthesize the voice or likeness of a minor unless a parent or legal guardian has validly authorized the specific lawful use and the content is age-appropriate.
• Do not use a voice, face, body, or other identifying characteristic to imply an endorsement, statement, performance, or affiliation that did not occur.
• Consent to one project or purpose does not imply consent to a materially different use, redistribution, sexual context, political message, or commercial endorsement.

Synthetic-media disclosure and provenance

When an AI-generated or materially AI-manipulated image, video, or audio could reasonably be mistaken for an authentic record of a real person, object, place, entity, or event, clearly and conspicuously disclose that it is synthetic or AI-manipulated before or at the time it is shown. Deepfakes, public-interest material, endorsements, testimonials, advertising, and realistic news-style media require particular care.

For artistic, fictional, satirical, or analogous works, a disclosure may be made in a manner appropriate to the work, but it must remain reasonably noticeable and may not be used to facilitate deception or violate another person's rights.

Outputs may include visible labels, metadata, watermarks, credentials, or other machine-readable provenance supplied by Sundream or an AI provider. You may not intentionally remove, conceal, falsify, or circumvent a provenance marker or safety label when doing so would mislead people or violate law. If an editing or distribution platform strips a marker, you remain responsible for a clear disclosure wherever one is required.

Intellectual property and privacy

• Do not upload, reproduce, or generate material intended to infringe copyright, trademark, publicity, privacy, contractual, confidentiality, or other rights.
• Do not expose private communications, credentials, financial information, health information, precise location, or other sensitive personal data without authorization.
• A model accepting a request or producing an output does not mean the request or output is lawful, original, accurate, cleared, or safe to publish.

Safety systems and review

Sundream and its AI providers may use automated prompt or output filtering, abuse detection, rate limits, account and generation signals, and limited human review to enforce this policy, respond to reports, secure the service, and comply with law. Our Privacy Policy describes related processing.

Safety systems can make mistakes and cannot detect every harmful request or output. You remain responsible for your content and use even when a request passes an automated check.

Enforcement

We consider severity, intent, context, actual or potential harm, account history, credible evidence, and legal obligations. We may block a request, remove or disable content, restrict a model, feature, API key, account, or workspace, require corrective action, issue a warning, preserve evidence, suspend access, or permanently terminate service. Severe or urgent violations may result in immediate action and, where legally required or necessary to address an imminent threat, a report to an appropriate authority.

Attempts to evade a safety control, use another account after enforcement, retaliate against a reporter, or conceal a violation are separate violations.

Reporting and appeals

Report suspected harmful AI media, impersonation, non-consensual content, voice or likeness misuse, threats, illegal activity, or other violations to support@sundream.studio. Include the relevant project, workspace, URL or file, the people or rights affected, and enough context for us to investigate. Do not send illegal sexual content involving minors; provide location and identifying details instead.

If we take action against your content or account and you believe it was a mistake, email support@sundream.studio with the subject “Safety appeal.” Identify the decision and explain the relevant context, authorization, consent, or corrective action. A person not responsible for the original decision will review an appeal where reasonably practicable. Repeated or abusive appeals may be closed without further response.

Changes and contact

We may update this policy as AI capabilities, abuse patterns, provider restrictions, and law change. We will update the date above and provide reasonable notice when a change materially affects user obligations. Material changes are handled under the reacceptance process in our Terms of Service.

Questions about this policy can be sent to support@sundream.studio.