
This announcement comes one month after OpenAI published a report identifying trends in the use of artificial intelligence by malicious actors, but without conclusive evidence of the use of current models in launching new attacks. However, the latest update from Google’s Threat Intelligence team confirms the exploitation of artificial intelligence in the development of adaptive malware.
Several new types of malware have been discovered that use artificial intelligence in innovative ways, including “FRUITSHELL,” a reverse shell specifically designed to bypass AI-powered security systems. There is also “PROMPTFLUX,” which leverages the Google Gemini API to dynamically rewrite its source code.
The discoveries also included “PROMPTLOCK,” a ransomware program that leverages artificial intelligence to generate and execute malicious scripts, in addition to “PROMPTSTEAL,” a data collection tool that uses artificial intelligence to create prompts for data theft. Finally, there is “QUIETVAULT,” which uses artificial intelligence to search for more secrets in infected systems.
The report noted key new trends, including the use of sophisticated social engineering techniques to bypass artificial intelligence security barriers, where attackers impersonate security researchers or students participating in security competitions. (zdnet)