Everything Dawnwood renders is AI-generated video, and every frame of it is traceable. This page describes the security mechanisms we operate to trace AI-generated content, the controls that keep real people out of it without their consent, how we label synthetic media when it leaves our hands, and the response plan we follow when something goes wrong. It applies to the console, the agent API, and the series we produce ourselves.
1. Traceability: every output has a record
- Per-call generation ledger. Each call to a rendering, image, voice or language-model provider is written to an append-only ledger with the provider, model, provider request id, timestamp, the account that requested it, the production it belongs to, and the prompt and reference assets that were sent. A finished clip can be walked back to the exact inputs and account that produced it.
- Reference assets are retained with the production. Character reference images, voice references and scene plates stay attached to the production record for as long as the production exists. If a likeness question is raised, we can show precisely which references were used and who uploaded them.
- Delivery packs carry provenance. Every delivery pack includes a machine-readable manifest of each file (measured duration, resolution, bitrate, checksums), a chain-of-title document, and an AI-content disclosure stating that the series is AI-produced.
- Retention. Generation records are kept for at least 12 months after the production's last activity, and longer when a report, dispute or legal request is open.
2. Preventing deepfakes and non-consensual likeness
- Fictional casts by default. Dawnwood characters are synthetic performers drawn from our actor and voice libraries. A production does not need, and normally does not contain, any real person's face or voice.
- Real people only with written consent; minors never, as set out in Acceptable Use Policy §1. Public figures are not an exception. Where a rendering provider offers an identity-verified real-person asset library, real faces are used only through it, so consent is bound to a verified identity rather than an uploaded file.
- Provider moderation is respected, never circumvented. Our rendering providers run input and output moderation, including likeness, celebrity and copyright checks. A rejected shot is retried only with the same inputs (moderation is probabilistic) or with a compliant replacement reference. We do not alter inputs to evade a moderation decision, and repeated rejections of the same face put the shot on hold for human review.
- No impersonation of real events. Content designed to pass as real footage of real people or events is prohibited (AUP §3) and is treated as a deepfake incident under section 4 below.
- Account controls. One account per person, staff access to customer media only through an audited admin surface, and least-privilege access to production storage.
2a. Real-name verification and violation handling for platform customers
- Every account is tied to a verified identity. Console and agent-API accounts are created through a verified email or Google sign-in, one account per natural person or company (AUP §4); paid plans are additionally tied to a payment identity.
- Real-person features require real-name verification. Any feature that renders a real person's face or voice is available only after the person depicted has completed identity (real-name) verification through our rendering provider's verified real-person asset flow, and only from an account that has accepted the portrait-authorization terms. Uploading a real face any other way is blocked by provider moderation and is a violation.
- Violation handling. Reports and detections are handled under section 4: triage within 24 hours, removal and account freeze within 24 hours of confirmation, evidence preserved for 90 days, and suspension, termination and credit forfeiture under AUP §6. Repeat or severe violators lose access to real-person features permanently.
3. Labeling synthetic media
- Series published through connected accounts carry each platform's native synthetic-media label wherever the API allows it (for example YouTube's synthetic-media flag and TikTok and Instagram's AI-generated labels).
- Delivery packs include an AI-content disclosure that licensees agree to preserve. Removing our labels or disclosures is a violation of the Acceptable Use Policy.
- Our own series are described as AI-produced on every public page that presents them.
4. Incident response plan
An incident is any report or internal detection that Dawnwood output depicts a real person without consent, impersonates a real person or event, sexualizes a minor, or otherwise violates the Acceptable Use Policy or applicable law.
| Stage | What happens | Target time |
|---|
| Report | Reports go to abuse@dawnwood.ai (copyright notices to legal@dawnwood.ai). The mailbox is monitored every day, including weekends. | Acknowledged within 24 hours |
| Triage | We identify the production, account and generation records involved using the ledger in section 1, and assess the report against this policy and the AUP. | Within 24 hours of the report |
| Containment | Confirmed violating content is removed from Dawnwood-hosted storage and from every connected channel we control, and the responsible account's rendering is frozen while the review completes. | Within 24 hours of confirmation |
| Preservation | Prompts, reference assets, generation records and the output itself are preserved as evidence for at least 90 days, and made available to the affected person or to law enforcement on a lawful request. | At containment |
| Notification | The reporter, and the affected person where identifiable, are told the outcome. Where a rendering or distribution partner's rules require it, we notify them as well. | Within 72 hours of resolution |
| Enforcement | Violations lead to suspension or termination and forfeiture of credits under AUP §6. Content involving minors is reported to the relevant authorities, not just removed. | With the outcome |
After each confirmed incident we record what allowed it and change the control that failed: a stricter default, a new automated check, or a policy clarification on this page.
5. Cooperation with providers and platforms
We comply with the usage rules of our rendering providers, including their real-person asset, copyright and portrait rules, and with the synthetic-media policies of the platforms we publish to. Lawful requests from platforms, rights holders or authorities are handled through legal@dawnwood.ai.
6. Contact
Abuse reports: abuse@dawnwood.ai · Legal and copyright: legal@dawnwood.ai · Dawnwood Inc. (Delaware, USA)