Research Reveals Security Concerns Surrounding ChatGPT

By Romy M May 14, 2024

In recent AI developments, researchers have uncovered potential security risks inherent in Artificial Intelligence (AI) tools such as OpenAI’s ChatGPT. Despite the rapid proliferation of AI tools from various companies, it has become evident that their utilization may be somewhat risk-free.

Image Courtesy of Miha Creative/Shutterstock

A deeper investigation into these groundbreaking technologies has revealed the potential susceptibility of users to various security threats despite the current absence of such issues. Notably, regulatory bodies have already raised concerns surrounding AI safety. Among the identified risks, researchers have highlighted the vulnerability of AI platforms like ChatGPT and Google’s Gemini, the latest recently launched version.

The research has identified a particular type of malware threat targeting the GenAI ecosystem, exemplified by a malware worm known as Morris II. Named after the infamous Morris worm of 1988, which caused widespread disruption by infecting a significant portion of internet-connected computers, Morris II exploits flaws in the architecture design of GenAI platforms rather than any specific vulnerability in the GenAI service itself.

Morris II operates by replicating and spreading throughout systems, often without requiring any user interaction. Unlike conventional GenAI operations that rely on user prompts and text-based instructions, Morris II manipulates these prompts, unknowingly coercing the GenAI into executing malicious actions.
To mitigate the risk posed by malware worms like Morris II, AI users are advised to exercise vigilance when handling emails and links from unknown or untrusted sources. Additionally, investing in robust antivirus software capable of detecting and removing malware is recommended as an effective defense mechanism against these threats.

Furthermore, implementing security measures such as using strong passwords, regularly updating system software, and minimizing file-sharing activities can help reduce the susceptibility to malware worms and other cyber threats.

Image Courtesy of Robalito/Shutterstock

Amid these security concerns, Sam Altman’s OpenAI introduced a new AI tool capable of replicating human voices. Known as Voice Engine, this tool requires only text input and a short 15-second voice recording to recreate a person’s voice. However, given its GenAI model, there is a significant potential for exploitation by malicious actors once it is fully deployed following the ongoing testing phase.