MOSCOW, May 13. Fraudsters have already learned how to use neural networks, adapt to the evolution of technology and actively use deepfakes, said Vladislav Tushkanov, a leading data scientist at Kaspersky Lab.
«
“The generative networks associated with images and voice can be used to synthesize evidence for something that didn’t exist—deepfakes. And they may well be used in fraud, for example. There have been many cases where scammers have created deepfakes to write social media posts with a link to a fraudulent page, one of the high-profile examples is the use of the image of Elon Musk,” he said.
At the same time, according to the expert, scammers are constantly evolving: people have already learned not to trust pictures, so attackers have started creating deepfake videos that are harder to distinguish.
«And it's not that deepfakes are bad technology. They can be used in training videos, they allow you to make them high-quality, lip-synching with speech when translating into different languages during dubbing, they are already widely used in cinema. But for fraudsters, this is including the opportunity to make people believe in their message,» Tushkanov believes.
The hype around neural networks also plays into the hands of scammers, the source said.
«The same ChatGPT is available under the web interface, but you can't access it from any country, they even stopped registrations for a while. Fraudsters actively use this. Recently we had a study: resources appeared on the Internet supposedly for downloading the ChatGPT client and still immediately get $ 50. Of course, it was not a client, but a Trojan, and there are enough such cases now, «he said.
However, according to Tushkanov, to ban artificial intelligence by law, as they did, for example, in Italy, is not an option.
«It is necessary not to regulate the technology itself, but to apply it correctly. Different countries have different rules. Somewhere they force you to warn that you are talking not with a person, but with a robot, sign the generated picture so as not to mislead people that this is a photograph» , he believes.
ChatGPT, an artificial intelligence chatbot developed by OpenAI (one of its founders is Elon Musk), gained popularity at the end of last year, gathering its first million users in less than a week. At the end of January, Microsoft announced that it plans to invest billions of dollars in OpenAI. And in March, the company introduced a new multimodal GPT-4 model that can recognize both text and images and solve complex problems with greater accuracy.
But the bot's imitation of a conversation has sparked speculation that it could crowd out professional writers and even threaten Google's core search business. In February, the media reported that the Chinese authorities banned the largest technology companies Tencent Holdings and Ant Group from providing customers with access to ChatGPT services. And at the end of March, Italy became the first Western country to restrict the work of ChatGPT due to violations in the collection of user information.