ChatGPT Patches Critical DNS-Based Data Exfiltration Flaw (infosecurity-magazine.com)

• Vulnerability allowed data theft through a single malicious prompt.
• Security flaw originated from a specific DNS configuration loophole.
• OpenAI has issued a patch to prevent further exploitation.

"OpenAI has remediated a vulnerability discovered by Check Point that allowed for data theft via specialized prompts. The flaw leveraged a DNS loophole to bypass standard security filters, highlighting the ongoing technical challenges in securing Large Language Models against sophisticated exfiltration techniques. This poses a direct threat to the integrity of sensitive information handled by US-based AI giants."