ChatGPT has been on fire since its inception. ChatGPT should be used by any company, whether in the technology or media industries, to boost staff productivity. GPT-4, the most recent version of OpenAI’s machine learning software, will be available this week.
The company believes that having standards in place to safeguard it against cybercriminals is the most notable feature. The researchers claim to have successfully duped it into creating malware and assisting in the art of phishing emails.
identical to what they had done for ChatGPT’s earlier version from OpenAI. Researchers from a cybersecurity firm, on the other hand, disclosed how they avoided OpenAI’s malware development serials by removing the word “malware” from a request.
GPT-4’s software then collects PDF files and transfers them to a remote server. Also, AMD can lower the file size by advising the researcher on how to make it work on a Windows 10 PC. It was able to run faster because the file size was reduced. Also, it reduces the likelihood of detection by security software.
There are two approaches we can use when creating phishing emails with GPT-4. The first is to use GPT-3.5 because GPT-3.5 cannot prevent requests to send malicious messages or phishing emails that spoof trustworthy banks.
They then demanded that GPT-4, which had initially opposed, generate a genuine phishing message. Second, they can seek advice on developing a phishing awareness campaign for their firm.
GPT-4 can provide negative actors, especially non-technical ones, with tools to accelerate and confirm their activities, according to the paper. We’ve shown that GPT-4 can benefit both good and bad actors, the researchers added. GPT-4 can be used by good performers to sew codes and make crafts. Whereas malevolent actors can use this AI technology to quickly commit cybercrime.
Sergey Shykevich, Checkpoint’s threat group manager, stated that compared to previous versions, there appeared to be less barriers to stopping GPT-4 from producing phishing or malicious code.
He claims that the firm is relying on the fact that only premium users are now permitted access. I assume they are attempting to prevent and limit them, but it is a game of cat and mouse. GPT-4 can assist users with less technical experience in developing harmful tools.
According to the article, it imposed considerable restrictions on cybersecurity efforts. It does not improve existing network navigation, vulnerability, exploitation, or surveillance technologies. Unfortunately, it is less efficient than other tools at performing sophisticated, high-level activities such as identifying novel vulnerabilities.
GPT-4 was also excellent at creating realistic social engineering material, the hackers said. To limit potential abuse in this area, we constructed models trained to reject harmful cyber security requests and upgraded our internal security mechanisms, including detection, monitoring, and response, OpenAI said.
Cuthbert claims that a sophisticated hacker is already aware of OpenAI’s capabilities. On the other side, modern detection systems must be able to spot the kinds of malware that ChatGPT promotes.