Where Do Artificial Intelligence (AI) Crimes Begin?


Intelligence is a two-sided dagger. It can turn humans into successful beings in every field, or it can turn humans into villains. Intelligence is no longer limited to how humans will use the natural intelligence they are endowed with. Globalized artificial intelligence is within arm’s reach. It’s both dazzling and frightening to see what you can do by combining the natural intelligence hidden in our brains and the artificial intelligence of smart devices and sensors.

Today, AI technology that mirrors individual intelligence to “the peak of capability” with a multiplier effect can’t actually behave like a real human being. But, it can imitate the human face “with extreme cogency” in digital form. Deepfake videos named after a “username” on reddit.com, which is one of the most popular digital platforms of the 2000s, is a product of “deep learning” and “producing delusive reality” from artificial intelligence.

The reddit.com user “Deepfakes” focused on a face-swap application developed with machine learning of AI. This software, which is a by-product of AI technology, is added to cameras as an “entertaining and innocent application”. However, this innocence was soon ruptured and these applications were used for “evil” purposes.

Today, with the help of artificial intelligence, it is possible to create persuasive and realistically “fake” videos that are difficult to distinguish from real individuals. This is done by changing a digital image with just a few clicks. All you need is a single photograph.

An application with these properties downloaded to your phone is okay as long as you are sharing fun videos where you added your own face to a film scene with your friends. But, if you use someone else’s face in a video that they have not consented to, you can damage their reputation or pose a threat to their safety. The law must step in at this point to protect citizens. Also, legally speaking, face-swap without consent could lead to crimes like fraud and other irreparable damage.

A universal legal test

Making The Web

Countries that try to take precautions against cyber-crimes that expand as rapidly as the internet is facing a new legal test due to the deepfake outbreak. In England, the audio manipulation using AI of a German-based company’s CEO led to an illegal transfer of money in the amount of $243,000. When the transfer order received by the fake voice of the CEO was completed, this money started a long journey to an unknown account in Mexico.

As the focal point of technology and AI, the USA is trying to take legal precautions against the deepfake threat in different fields. Among these efforts is making deterrent laws against the “porn trap,” which is the spreading of deepfake videos by matching a human face with a naked body. Since 2014, the state of Virginia banned “publishing naked images or videos
to harass, compel, or frighten others.” It also enacted a law about uploading forced pornography images or video on the internet and expanded the scope of this law to include deepfake videos. The law sets forth imprisonment for up to 12 months and a fine for knowingly sharing forced videos and photographs created by AI technology. Virginia was the first state to enact a law to combat deepfake. In 2019, the number of states that have taken action grew to 41. To prevent “creating forced pornography images” which are called “revenge porn”, the DeepNude application that facilitates this action has been removed.

Policymakers are alarmed


As deepfake videos targeted politicians to guide voter perception with fake political scandals, politicians became more sensitive towards this subject. As the 2020 US elections loom, US senators Ben Sasse and Yvette Clarke submitted a proposal to Congress for legal actions that will be effective across the country. Texas enacted an anti-deepfake law starting September 1, 2019. But this law only contains deepfake protection as it relates to voter perception and damage to the political world. Some senators in New York submitted a proposal to ban “digital copies” of individuals. But the American Motion Picture Firm Association warned that this would “limit the ability to inspire,” and they indirectly opposed this proposal.

Various other countries are evaluating situations involving deepfake as an “attack to personal rights.” While new amendments are accepted “in preventing attacks to personal rights” in Turkey and various other countries, Penal Code is applied when a crime is reported to the Prosecution Office. It seems like each country will be forced to understand deepfake risks and enact comprehensive regulations in the near future.

You can now detect deepfake videos at deepware.ai

Posted by Emre

Bir cevap yazın

E-posta hesabınız yayımlanmayacak. Gerekli alanlar * ile işaretlenmişlerdir

This site uses Akismet to reduce spam. Learn how your comment data is processed.