Openai on Friday disclosed What he banned the set of accounts that used his Chatgpt tool to develop a suspicious artificial intelligence tool (AI).
It is said that the social media is likely to listen to social media Models LamaWith the accounting records that use AI models to create detailed descriptions and analysis of documents for the apparatus capable of collecting data in real time and report on anti-Chinese stocks in the West and share ideas with the Chinese authorities.
The company was named an expert review of the ” -in -the -” behavior of the network in promoting and examining tools of observation ” – noted researchers Ben Nimma, Albert Zhang, Matthew Richard and Nathaniel Hartley, adding the tool designed to absorb and analyze messages and comments from the platforms, X , Facebook, YouTube, Instagram, Telegram and Reddit.
In one of the cases indicated by the company, the actors used the chat for debugging and changing the source code, which is believed to launch a monitoring software called “Qianyue abroad for public opinion AI Assistant”.
In addition to using its model as a research tool for publicly available information about analytical tanks in the US, as well as government officials and politicians in countries such as Australia, Cambodia and the United States transfer and analyze the screenshots of English -language documents.
Some images were announced protests to Ugur in different western cities and were probably copied from social media. It is now unknown whether these images were valid.
Openai also stated that he thwarted several other clusters who were abusing chat for various malicious activities –
- Scheme of deceptive employment – Network from North Korea associated with The deceived IT -Working Scheme This participated in the creation of personal documentation for fictitious entrants, such as resumes, online jobs and accompanying letters, as well as convincing answers to explanation of unusual behavior such as avoidance video calls, access to corporate systems from non -sanctified countries or work hours or work . Some of the fictitious applications were distributed to LinkedIn.
- Sponsored dissatisfaction – The network is probably Chinese origin, which participated in the creation of social media content in English and long articles in Spanish, which were critical of the United States, and later published Latin American news sites in Peru, Mexico and Ecuador. Some activities intersect with the famous cluster of activity Spamp.
- Scam for romance – Network of accounts that participated in the translation and generation of comments in Japanese, Chinese and English Romance and investment scams.
- Iran’s impact Nexus -The network of five accounts that participated in the generation of X and articles that were pro-Palestinian, about-gams, and pro-Iran, as well as anti-Israel and anti-ZSH, and divided into sites related to Iranian influences as surgery as International Union Virtual Media (Iuvm) and Storm-2035. One of the prohibited accounts was used to create content for both operations, which indicates “previously not reported relationships”.
- Kimusuki and blanoraf – Network of accounts managed by North Korean threatening subjects related When collecting information related to cyber -instruments and Topics related to cryptocurrencyand rehabilitation code for the remote desktop (RDP).
- Youth Initiative Hidden Surgery – A network of accounts that participated in the creation of English -language articles for the site called “Ghana’s Enlargement” and comments on social media aimed at the Ghana presidential election
- Affer task – A network network is probably infections It attracts people who do not suspect, to workplaces that perform simple tasks (such as love video or write reviews) in exchange for receiving a non -existent commission, which requires the victims to say goodbye to their own money.
Development comes when AI’s tools are increasingly used by bad subjects to facilitate Companies misinformation with cyber-with support and other harmful operations.
Last month Google Degrate Intelligence Group (Gtig) disclosed More than 57 different threats associated with China, Iran, North Korea and Russia used their chat -bot Gemini AI to improve several stages of the attack and research in current events, or performing content, translation and localization.
“Unique ideas that AI companies can get threatens from subjects, especially valuable if they are divided with suppliers up the course, such as hosting and software developers, distribution platforms down on the current such as social media and researchers Open source, ”Openai said.
“Equally understanding that suppliers and researchers up the current and down the current open new ways to detect and execution of AI companies.”
