AI Model Representation & Ethical Use Policy

Definitions

For the purposes of this Policy, the following definitions apply:

1Scope of AI Model Activities

By signing this annex, the agency confirms that it:

• Represents, creates, or collaborates with AI-generated models, or works with real models in AI-related engagements — currently or in the future — including digital alterations, digital twins, voice synthesis, and video or animation content
• Understands its obligations under the IMF Ethical Use Policy for AI Models
• Commits to executing all AI-related engagements involving a Real Model in accordance with the AI Framework, and to using Framework Aligned Consent Records — typically generated via Model ID's AI Consent Tool — for such engagements. The AI Framework governs the substantive protections applicable at the per-engagement level; this Policy governs the agency's standing obligations as a certified party.

2Digital Twin Ownership

• Any digital twin or AI-generated representation that replicates the likeness, voice, biometric data, or identity of a real model is the sole property of that individual. This includes exclusive rights to consent to, control, monetize, revoke, or restrict any use, reproduction, distribution, or adaptation of the digital twin across all platforms and media.
• Creation, licensing, or commercial use of a digital twin may only proceed with the model's explicit, written, and informed consent, documented in accordance with IMF standards. Models may withdraw consent for future use at any time.
• Profits derived from digital twins must be transparently reported and equitably shared with the real model, with no hidden exploitation or impersonation
• Digital twins can be registered to the real model's verified identity via Model ID's Digital Twin Authorization service. Registration is recommended to establish public, verifiable ownership so counter-parties can identify the authentic version of the twin. This service is separate from and independent of Model ID's AI Consent Tool. Agencies, brands, and AI creators are prohibited from developing or deploying digital twins without valid ownership confirmation by the real model.

3Ethical Use of Non-Human AI Models

• Agencies working with non-human AI models (fully synthetic) must ensure transparency in client briefs and label such models clearly in public-facing material
• Use of AI to deceive, mislead, or obscure real model participation is prohibited
• Synthetic models must be clearly distinguishable from real individuals in all commercial contexts

4Model ID Integration

• Real models can register their digital twin via Model ID's Digital Twin Authorization service — a model-initiated verification that establishes public ownership of the twin, allowing counter-parties to identify the authentic version. Registration is recommended but not mandatory, and operates independently of Model ID's AI Consent Tool (described below).
• Agencies are responsible for maintaining accurate usage records for all AI model engagements and providing them to the relevant model upon request. Model ID's AI consent contract tool enables agencies, models, and brands to formally document per-project AI usage rights, with all parties receiving signed copies
• All digital representations must be linked to verifiable ownership records

5Consent & Licensing

• Written, informed consent is required for every use of a digital twin. Model ID's consent contract tool provides a structured 3-party workflow: the agency initiates the contract specifying AI usage type, duration, territory, and compensation; the brand/client signs; and the model gives final approval
• Consent must be specific to the usage type, duration, territory, and medium
• Consent is revocable — models and agencies may withdraw permission for future use at any time through the Model ID platform, with all parties notified automatically
• Blanket consent for all future AI uses is not permitted. Each project requires its own consent contract specifying the exact scope of AI usage

6AI Model Representation Disclosure

• AI models must never be presented, offered, or invoiced as if they were real models. When an AI model is proposed for a booking, this must be disclosed to the client in writing at the point of offer — not after commercial terms are agreed.
• For composite campaigns, the share of AI vs. real models must be disclosed
• AI-generated content must be labelled using available metadata standards (e.g., C2PA, IPTC) and visually disclosed where the content is published

7Misuse & Enforcement

• Agencies must not knowingly participate in, facilitate, or ignore misuse of AI-generated models — including but not limited to: non-consensual deepfakes (criminalised under French Criminal Code Art. 226-8 and equivalent provisions in other jurisdictions); use of a model's likeness, voice, biometric data, or motion data outside consented scope; deployment or licensing of a digital twin without valid model consent; distribution of AI content without required disclosure under the EU AI Act (Art. 50) or applicable law; and any impersonation, identity confusion, or reputational harm to the model.
• Substantiated misuse triggers immediate investigation under IMF's Enforcement & Compliance procedures (the Progressive Response Framework, published at model-id.com/certification/enforcement) and may result in suspension or revocation of IMF certification. Independently, and in accordance with the AI Framework's Liability & Penalties principle, the affected model retains the right to (a) publicly disclose the breach — with confidentiality obligations under any consent record waived to the extent necessary — and (b) report the breach to competent regulators, including data protection authorities under GDPR Art. 79 and market surveillance authorities under the EU AI Act.
• Agencies have a proactive duty to report to IMF any known or reasonably suspected instances of unauthorized digital twin creation, unauthorized AI use of certified members, or breach of the AI Framework by a third party — regardless of whether the agency itself is a party to the underlying engagement.
• Use of a certified model's likeness, voice, biometric data, or motion data as training or fine-tuning data for any AI system is prohibited. This applies whether such use is direct (by the agency or engaging party) or indirect (as a side effect of AI-tool terms of service). Engagements involving training data must be effectuated through separate legal arrangements outside this Policy and outside Framework Aligned consent records.

8Minors

AI-related work involving a Real Model who is under the age of 18 requires additional protections beyond those set out in this Policy. Certified agencies must comply with the IMF Underage Model Protection Protocol in full when executing any AI work involving a minor, and must not initiate any AI-related engagement — including creation, licensing, or use of a digital twin — for a minor except where the additional consents and safeguards required under that protocol are documented and in place. Where the Underage Model Protection Protocol imposes stricter standards than this Policy, the stricter standard prevails.

9Version, Effective Date, and Amendments

This Policy is Annex C to the IMF Agency Certification Agreement. It carries a version number and effective date shown at the top of this page. IMF may amend this Policy from time to time; amendments take effect on the effective date of the revised version. Certified agencies will be notified of material amendments and given a reasonable period (typically thirty days) to review and, where required by the nature of the change, re-attest. Continued conduct of AI-related work after the effective date of a revised version constitutes acceptance of the revision, subject to any express opt-out mechanism IMF may provide. IMF publishes the current and prior versions of this Policy at model-id.com/legal/ai-model-policy and model-id.com/about/policies.