Researcher Tricks ChatGPT Into Building Undetectable Steganography Malware

Using only ChatGPT prompts, a Forcepoint researcher convinced the AI to create malware for finding and exfiltrating specific documents, despite its directive to refuse malicious requests.