In a troubling development, cybersecurity experts have identified a series of malicious custom GPTs that exploit ChatGPT’s capabilities to serve as delivery mechanisms for Remote Access Trojans (RATs). This alarming trend emerged in late September 2023, as threat actors increasingly leveraged legitimate domains from OpenAI and Google to deceive unsuspecting users into downloading harmful software.
Context: Understanding the Threat Landscape
The rise of AI-driven technologies has revolutionized the digital landscape, offering countless benefits while simultaneously presenting new vulnerabilities. Custom GPTs, which allow users to tailor AI models for specific tasks, have gained popularity for their versatility and ease of use. However, this customization feature has also opened doors for malicious actors.
According to a report by cybersecurity firm Cybereason, the use of legitimate domains to host malicious content is a tactic designed to bypass security measures and build trust with potential victims. This approach has been noted in previous campaigns, often referred to as ClickFix-style attacks, which take advantage of user familiarity with trusted brands.
Detailed Coverage: Analyzing the Malicious Campaign
Cybersecurity analysts discovered that these malicious custom GPTs mimic legitimate applications, making them appear harmless. Users seeking to enhance their productivity or engage with AI tools are lured into downloading and executing these RATs, which can grant attackers unauthorized access to their systems.
The campaign reportedly began on platforms popular among developers and tech enthusiasts, where discussions around custom GPTs are prevalent. Threat actors have been sharing links to these malicious tools under the guise of legitimate software enhancements. Once installed, the RATs can stealthily record keystrokes, capture screen images, and even exfiltrate sensitive data.
Expert Perspectives: Insights from the Cybersecurity Community
Experts warn that the sophistication of these attacks is increasing.