How Effective is Regulatory Scrutiny in Combating Deepfake Pornography?

Since 2023, it’s evident that deepfakes are the most concerning cyber threat currently prevailing and continuously evolving at a high pace due to AI advancements. Several challenges arise when we discuss criminalizing the creation of non-consensual deepfakes, especially deepfake pornography that targets innocent people including children. A few commendable efforts by the regulatory authorities have been observed in curbing this threat since 2019. But is it enough and has it visibly lowered the illicit use of deepfakes or are regulators just pretending to act responsibly?

Today we will try to find answers related to deepfake regulations, and mitigation strategies taken by authorities, and let’s try to expose, curtail, and bring deepfake creators to justice.

Tech Overview of Deepfakes

Deepfakes were intended to do good but eventually, this extremely precise graphic technology fell into the wrong hands. Deepfakes are created by a group of imposters using highly sophisticated image manipulation application software employing GAN (Generative Adversarial Network) model training or CNN (Convolutional Neural Network) training. Both these techniques supplement the creation of AI-based deepfakes that are the most realistic and highly convincing having the ability to spoof biometric identity verification systems too. Deepfakes are created for almost every visible and easily identifiable biometric trait of human beings.

Deepfakes: Biometric Breakdown

Deep damage Due to Deepfakes

There is a list of deepfakes being used for negative activities. Most concerning is the creation of deepfake pornography that damages the social life of the victims if the explicit content goes viral. Secondly, deepfakes are used to spoof identity verification systems using biometrics, particularly using facial deepfakes. Look at the graph below, which shows how deepfakes creation spiked recently.

Statistics from 2023 onwards show that the biggest attack vectors that deepfakes target are photos and videos on screen. This refers to the following three categories of facial images:

• Photos on Screen used to create AI images (deepfakes) account for more than 56% of the Selfie Images uploaded/shared on social media and people’s smartphones.
• Video on screen used to create deepfake videos accounts for more than 26% of the shared videos on different platforms like Instagram, Facebook, etc.
• In the motion category, the video on screen accounts for more than 80% of deepfake creation threat vectors.

These statistics show the vulnerability of user’s facial data that has gone viral in both picture and video formats over the internet facilitating abundant facial data for deepfake makers. Therefore, a strong legislative framework is needed to curb deepfake creation.

Who is Fighting Deepfakes?

The US’s regulatory bodies in different states have made the most impactful progress against illicit use of deepfake technology. As of now, no robust legislation or a deepfake law that is robust enough for which we can say that it is combatting AI deepfakes in real time criminalizing them and punishing the culprits. Out of 10 laws in the US against deepfakes, 6 impose criminal charges and the rest are just on the books. However, there are a few acts listed below that regulate the use of GAN, Machine Learning, and AI’s use for creating illicit material to some extent.

Identifying Outputs of Generative Adversarial Networks Act – December 2020

It mandates the National Science Foundation to fund the R&D and standardizing of GAN outputs and similar technologies.

Deepfake Report Act – 2019

It requires the Directorate of Science & Technology in the US Department of Homeland Security to report digital content forgery at required intervals

Deepfake Accountability Act – September 2023

This act aims to safeguard national security against the potential threats of deepfake technology and provide legal backing and assistance to the victims of damaging deepfakes.

Defiance Act – 2024

It is enacted to ensure and improve the rights to relief for victims of non-consensual activities involving digital manipulation of content primarily for illicit gains and other purposes.

Protecting Consumers from Deceptive AI Act – 2024

It mandates the National Institutes of Science & Technology (NIST) to develop technological standards and SOPs including guidelines to detect Gen-AI content. This will ensure that the Gen-AI created/manipulated audio or video must not go undetected. It should include a disclosure acknowledging the use of Gen-AI to generate content for other purposes.

Statewise Legislations Against Deep Fakes

Are These Deepfake Legislations Sufficient?

A blunt answer to this question is no. The reason is that regrettably, deepfakes are spreading like a virus everywhere targeting innocent women, children, and even men. Cybercriminals seek different gains from deepfake images and videos including non-consensual pornography, child pornography, cyberbullying, and extortion through blackmailing. Terrifying statistics reveal that 75% of deepfake pornography viewers don’t even feel guilty about it. It costs $0 and only 20 minutes to make a 1-minute deepfake video that is highly realistic and the number of deepfake nudes is increasing every minute.

This shows that the above legislation is on paper and has yet to be visibly effective. However, continuous efforts to create mass user awareness, the efficient use of facial liveness detection, and the removal of explicit deepfake material are good starts. Regulators should support non-profit organizations that are supporting the victims of deepfakes. Moreover, strict legal actions including imprisonment, fines, and a global crackdown against deepfake pornographic websites should be enacted at once. 

A critical role can also be played by face recognition tool providers who can support law enforcement agencies and individuals in detecting their deepfakes online and the rapid removal of such data. Users can reach out to FRT providers and law enforcement with complaints and can secretly remove such data to prevent its spread further.