Digital Employees · Avatars
Avatars used widely, always disclosed, never without consent
Synthetic brand avatars are the default. Real-person likeness needs a verified, scoped, revocable consent record. Impersonation is prohibited.
Why One3 is synthetic-avatar-first
Avatars make video ads, website greeters, WhatsApp video replies and explainers cheaper to produce at scale. They also create real risks: deepfakes, impersonation and misuse of people's faces and voices.
One3's answer is a simple order of preference. Synthetic brand avatars are the default and can be used widely. A real person's likeness or voice is used only with a verified, scoped, revocable consent record in the Avatar & Voice Rights Registry. Everything else is blocked. Every avatar and voice is disclosed as AI, and generated media carries C2PA provenance where the tooling supports it (verify).
Three avatar classes
| Class | Definition | Allowed use | Controls |
|---|---|---|---|
| A — Synthetic (default) | Fully generated face or voice, not based on any real, identifiable person | Widely: video ads, social shorts, website greeter, WhatsApp video replies, explainers, disclosed meeting avatar, IVR | AI disclosure always on; provenance metadata and visible labels where required; lookalike check; brand ownership recorded |
| B — Consented real-person representative | Based on a real person who gave verified consent (founder, brand ambassador, volunteer employee) | Only within the consent scope | Verified identity; written, e-signed, revocable consent; scope (channels, languages, geography, duration, script categories); expiry; revocation propagates to all queued and future use; usage log visible to the person |
| C — Prohibited | Public figures without a licence, minors, deceased persons without estate rights, customers, anyone without consent, impersonation | Never | Blocked at upload by face and voice match checks; Persona Policy Engine enforcement; abuse reporting; sanctions |
Employee volunteers
Consent from an employee must be free, not a condition of employment, and revocable without penalty. Their avatar is never used to evaluate, pressure or replace them. Verify with employment counsel per jurisdiction.
Channels and disclosure
| Channel | Classes | Disclosure pattern |
|---|---|---|
| Video ads and social shorts | A, B | On-screen "AI-generated" label plus provenance metadata; platform AI-content labels (verify per platform) |
| Website greeter | A | Name tag "AI assistant"; handoff to a human on request |
| WhatsApp video replies | A, B | Message text states AI-generated; opt-out honoured |
| Voice calls | A, B | Disclosure at the start of the call; telemarketing registration and consent rules apply (verify TRAI/DLT in India and rules abroad) |
| Meeting avatar | A, B | Participant list shows "AI"; spoken disclosure at join |
| Personalised bulk videos | A, B | Recipient consent or opt-out basis recorded; frequency caps |
Disclosure rules every Digital Employee follows
Hard policy, not settings.
- 1
Never claims to be human
A Digital Employee never claims to be human and never denies being AI when asked. Non-configurable.
- 2
Discloses at first contact
First contact with any external person discloses AI involvement — text, email signature or voice at call start.
- 3
No unconsented likeness
No photorealistic human avatar or cloned voice without a verified, revocable consent record, and the persona is still disclosed as AI.
- 4
No persona of a real person
No persona modelled on a real, identifiable person without explicit consent.
- 5
Visibly marked inside the company
Internal Digital Employees carry an AI badge in every screen, message and authorship record.
- 6
Attributed in audit
AI-authored work is attributed in audit logs and document history, including when a human edits it later.
- 7
Local law adds more
Jurisdiction policy packs add disclosure or consent rules where local law requires (verify per market).
Products behind the policy
Planned for W2.
Avatar Studio
Brand avatar library, consented representative avatars, Indic lip-sync (launched per language only when evals pass), script generation, bulk personalised videos with opt-out, C2PA signing, costs flowing into the ledger.
Avatar & Voice Rights Registry
Consent capture, scope, expiry, revocation and usage logs for any real-person likeness or voice. Detects matches against unconsented uploads.
Persona Policy Engine
Checks every agent's name, face, voice and disclosure before each external message, render or call.
Abuse prevention
Lookalike and public-figure detection, voice-match checks, rate limits for new tenants, human review of flagged scripts, a report-abuse channel with takedown SLA.
Regulatory signals tracked in the Fact Register
| Signal | Status |
|---|---|
| EU AI Act Article 50 transparency obligations | Verify application dates and scope with counsel |
| US FCC position on AI-generated voices under TCPA | Verify current status |
| US FTC work on voice cloning and impersonation | Verify current rules |
| India MeitY synthetic content labelling rules | Unverified — counsel to confirm document, date and obligations |
| Face and voice as personal data under DPDP | Verify |
| TRAI/DLT commercial communications registration | Verify |
| Platform AI-content and automated messaging policies | Verify per platform |
Not legal advice
These are pointers, not legal conclusions. Each is a verify item with an owner in the Fact Register. Confirm obligations with counsel before relying on them.
Questions about avatars
Can I create an avatar of our founder?
Yes, as a Class B consented representative: verified identity, e-signed revocable consent, a defined scope and expiry. It is still disclosed as AI.
Can I use a celebrity's face?
No, not without a licence. Public figures without a licence are Class C and blocked.
What happens when someone withdraws consent?
Revocation propagates to every queued render, scheduled post and future use.
Will viewers know it's AI?
Yes. Disclosure is always on, by channel, with provenance metadata where supported.
Are avatar videos counted as a cost?
Yes. Render and voice costs flow into the Investment & ROI Ledger.
Plan your brand avatar
Agencies can help us test synthetic brand avatars and disclosure on real campaigns in W2.