Synthetic content marking policy

This Policy describes how ROIFORCIO GmbH marks the outputs of the Pitch Avatar service so that they are detectable as artificially generated or manipulated, in accordance with Article 50(2) of Regulation (EU) 2024/1689 (the “AI Act”). It forms an integral part of the Terms of Use.

1. Scope

1.1. This Policy applies to the outputs of the Service which constitute synthetic audio, image, video or text, including AI presenter videos and voice-overs, AI avatars, outputs of voice cloning, and outputs of video translation and dubbing.

1.2. ROIFORCIO acts as the provider of those AI systems. The obligations of the person who publishes or distributes the resulting content — the deployer — are set out in Section 4 of this Policy and in Section 5 of the Terms of Use.

2. Marking standard

2.1. Outputs generated by the Service carry a machine-readable marking identifying them as artificially generated or manipulated. The marking is applied at the point at which the output is produced or exported through the Service.

2.2. ROIFORCIO selects the marking techniques applied to each type of output — which may include provenance metadata, embedded content credentials, cryptographic signatures or watermarking — on the basis of the state of the art at the relevant time, the type of content concerned, the generally accepted technical standards, the specificities and limitations of the available techniques, and the cost of implementation, with the objective that the solution is effective, interoperable, robust and reliable to the extent technically feasible.

2.3. Markings are designed to remain detectable after the transformations to which content of the relevant type is commonly subject, such as re-encoding or compression, to the extent technically feasible. No marking technique is resistant to every transformation or to deliberate removal, and ROIFORCIO does not warrant that a marking will survive processing by third-party software or a determined attempt at removal.

2.4. The techniques applied evolve with the state of the art and with the applicable technical standards. ROIFORCIO may replace or supplement a technique provided that the objective set out in paragraph 2.2 continues to be met, and reviews this Policy accordingly.

2.5. Removing, altering, suppressing, degrading or circumventing a marking applied by the Service is prohibited under the Terms of Use and under the Acceptable Use Policy.

3. Third-party generative components

3.1. Certain generative capabilities of the Service are performed by third-party general-purpose AI providers acting as subprocessors of ROIFORCIO and identified in the Privacy Policy.

3.2. Where generation is performed by such a provider, ROIFORCIO ensures that the output finally delivered through the Service carries the marking required by this Policy, whether by relying on the marking applied by the provider or by applying its own marking layer at the point of export. The obligations of those providers at the level of the underlying model remain their own.

4. Visible disclosure of deep fakes

4.1. In addition to machine-readable marking, the Service supports a visible indication that the content has been artificially generated or manipulated.

4.2. Where content generated through the Service constitutes a deep fake within the meaning of Article 3(60) of the AI Act, the deployer shall disclose that fact to the audience in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure, in accordance with Article 50(4) of the AI Act. Where the content forms part of an evidently artistic, creative, satirical or fictional work, the disclosure is limited to what is appropriate so as not to hamper the display or enjoyment of the work.

4.3. ROIFORCIO provides the technical means enabling that disclosure but does not assume the deployer’s legal responsibility for making it.

5. Exceptions

5.1. The machine-readable marking obligation does not apply to the extent an AI system performs an assistive function for standard editing, or does not substantially alter the input data provided by the deployer or the semantics of that input data.

5.2. The obligation does not apply where the use of the AI system is authorised by law to detect, prevent, investigate or prosecute criminal offences.

6. Verification and complaints

6.1. A person who wishes to verify whether content originates from the Service, or who considers that content generated through the Service infringes their rights, may write to info@roi4cio.com. ROIFORCIO examines such requests and, where the necessary information is available to it, confirms whether the content carries a marking applied by the Service.

6.2. ROIFORCIO does not retain a register of generated content permitting the identification of the author of an arbitrary item of content beyond the data it holds under its Privacy Policy and Data Processing Agreement.

7. Applicable dates

ROIFORCIO keeps under review its adherence to that Code of Practice and to the harmonised technical standards adopted under the AI Act, and updates this Policy accordingly.

8. Changes

8.1. This Policy may be amended in accordance with the procedure set out in Section 13 of the Terms of Use.