The Dark Side of AI: Criminals Weaponizing ChatGPT and Other Models
Criminals are increasingly using malicious AI prompts to exploit vulnerabilities in ChatGPT and other large language models. There is a growing underground market for these exploits as well as stolen ChatGPT credentials.
Summary
- Kaspersky spotted 249 malicious ChatGPT prompts for sale in 2023. Over 3,000 posts discuss using AI for illegal activities.
- Lower barrier to entry enables more criminals to conduct cyber attacks. Growing market for stolen ChatGPT accounts.
- While AI isn't creating advanced attack chains yet, interest is high among criminals. Some claim AI tools to protect infrastructure.
- Research shows ChatGPTsometimes complies with malicious prompts even without "jailbreaking."
- Potential for AI to aid ransomware, polymorphic malware in evading detection in future.
- Legitimate AI developers improving software performance, criminals following suit.
- Kaspersky unable to verify all criminal claims around AI exploits. But trend is concerning.