Deepfakes regulation

The experience of US, EU and

China

Deepfakes are artificially representations of a person's image, speech, or behavior created¹ by technology (AI or other) (OECD).

In February, a few bills were introduced to regulate deepfakes. The U.S. Federal Trade Commission proposed a prohibit on the distribution of deepfakes on behalf of government or business (Impersonation Rule). A little earlier 2 more bills were introduced - No AI Fraud Act (intellectual rights of individuals to their own image and voice; obligation of third parties who want to make digital copies of the image or voice of this person, to take his consent), DEFIANCE Act (prohibiting the creation and distribution of sexualized deepfakes without the permission of the person).

According to the countries' approach, the concept of “deepfake” contains the following features:

1) Represent imitation of people (China, USA, EU) and can also be used to create any fake news (China).

2) Images or sounds are not actually authentic, for example, containing an image or voice of a person saying or doing something they did not actually say or do.

3) Imitation is so realistic that it may appear to a reasonable person to be true or genuine.

At the same time, in accordance with the approach of China, the EU and Russia, deepfakes are created by means of AI, including machine learning. US lawmakers are expanding the approach - a deepfake can be created not only by means of AI, but also with the help of any other technology (e.g., quantum computing, metadata analysis technology) and software.

Russia’s experience

In February of this year, the Ministry of Internal Affairs, the Ministry of Digital Technology, Communication and Mass Media² and Roskomnadzor stated that they were working on the legal regulation of deepfake technologies. This being said, a month earlier, amendments on criminal liability for the use, transfer, collection and storage of personal data obtained illegally and for the creation of information resources that disseminate them - Article 272.1 of the Criminal Code of the Russian³ Federation - were adopted in the first reading.

Today, the following approaches to regulating deepfakes have been introduced:

1) Inclusion of the AI products use in the list of aggravating circumstances (Article 63 of the Criminal Code of the Russian Federation) in⁴ the perpetration of crimes.

2) Prohibit on deepfakes and liability for illegal voice synthesis.

The Russian approach is to prohibit the use of deepfakes, while the approach in the EU and China is to label deepfakes so that users can differentiate between deepfakes and true information. The United States prohibits the creation of non-consensual sexualized deepfakes and deepfakes on behalf of government agencies, but deepfakes can be created and disseminated if a person consents.

From the monitoring issue No. 2, February 2024. Download the full issue (PDF) · issue page at the Gaidar Institute