OpenAI warns hackers are using AI to operate more sophisticatedly

Hacking groups around the world are exploiting AI to compose malware, create phishing emails and harvest data, making cyberattacks harder to detect and prevent, OpenAI warns.

OpenAI's October report found that cyber actors globally are primarily integrating artificial intelligence (AI) into existing attack pipelines to improve performance, rather than developing entirely new attack tools or methods.

According to OpenAI, much of the activity detected involved using AI to assist with familiar tasks like developing malware, building control infrastructure, creating more sophisticated phishing emails, and gathering information about targets.

“The blocked teams were mostly incorporating AI into existing processes, rather than building new processes based on AI,” the report states.

The report also found that many user groups have leveraged large language models (LLMs) to mine knowledge in a variety of technical domains, from application development to network system configuration, often in a way that specializes each account for a specific task.

Several other user networks have been identified as using AI platforms to generate online content, including articles, images and videos , to serve media campaigns or promote certain viewpoints. However, most of these accounts are ineffective, with very low engagement levels.

In addition, OpenAI has also recorded the phenomenon of fraud centers and cybercriminals using ChatGPT to support online fraud activities in Myanmar and Cambodia. Some cases show that AI is used not only to edit content, but also to manage work schedules, human resources and finances.

However, OpenAI notes that AI is also being actively used to prevent fraud. It is estimated that ChatGPT is used three times more often to identify and avoid scams than to commit fraud.”

Most malicious requests are rejected by the system, but the report notes that there is still a “grey area” where technical or research-based requests can be exploited for unintended purposes.

However, OpenAI asserts that the code or content created is not malicious until it is misused outside the platform./.

(TTXVN/Vietnam+)

Source: https://www.vietnamplus.vn/openai-canh-bao-tin-tac-loi-dung-ai-de-hoat-dong-tinh-vi-hon-post1068941.vnp