Thousands of Dark Web Posts Expose ChatGPT Abuse Plans

Kaspersky reports observing over 3,000 dark web posts in 2023 discussing the abuse of ChatGPT and other LLMs, including jailbreaks, sales of stolen ChatGPT accounts, and the development of AI-enabled tools like FraudGPT and WormGPT to aid phishing, polymorphic malware, and deepfakes. Threat actors are sharing prompt sets and techniques to bypass safeguards and automate criminal tasks, lowering barriers to entry and complicating detection. The piece underscores growing AI-driven cybercrime activity and urges investment in AI-powered defenses and user vigilance, while noting the absence of confirmed malware leveraging some proposed methods to date.