*Per Ricardo Martins
The new version (4.0) of GPT Chat, a prototype of an artificial intelligence chatbot developed by OpenAI specializing in dialogue, is now available for those using GPT Plus chat and brings major updates to the AI. However, information security specialists draw attention to the provision of data, which has been improved in the new version, making it easier for cybercriminals to access files of interest.
Introduced on March 14, GPT-4 will have over 100 trillion parameters. The previous model (GPT-3.5) was trained with 175 billion parameters, that is, the tool will be able to create even more human-like content. With that, now, in addition to text, it is possible to present images, graphics and infographics to the AI. There was a lot of questioning about chat responses, but now version 4.0 offers 40% more factual responses than the previous version (3.5).
However, the update also has its downside. In addition to its limitations, such as the fact that its base only collects data from 2021 and the recurring user overload, experts in the field warn of the greater ease of access that cybercriminals will have with the new version. Much has been said about the GPT Chat and its benefits, the real estate and educational sectors, for example, have been using it massively for the production of content. Despite the positives, the technology has also been used on a large scale by cybercriminals.
With the popularity of the platform, scammers also started to use the bot to look for new ways to commit crimes, and now the update improves the data supply. People who don't have much knowledge about it can use the bot to produce malicious code.
Theft of files of interest, installation of tools that allow access door release (backdoor), sending infected malware, locating fake profiles on dating sites and reactivating idle conversations, expanding the use of phishing, are some of the actions of cybercriminals with the help of chat.
OpenAi even created a barrier to prevent hackers from using the bot for this purpose, but many criminals have already managed to bypass the system. Despite this, some IT directors claim that chat will increase cybersecurity through an optimistic view that technology and research professionals will gain far more capabilities from AI than cybercriminals.
Regardless of use, just as it is possible to optimize activities or create new threats, it is important to pay attention and always be up to date and protected, to avoid exposing new vulnerabilities.
*Ricardo Martins, information security analyst at Compugraf
EN 
PT 
