A simple 'stop' button click can exploit safeguards in ChatGPT or Microsoft Copilot, exposing unfiltered outputs. #ai #cybersecurity #aisafety #openai #microsoft #chatgpt #llms

https://buff.ly/3Bc8uzC

https://files.mastodon.social/media_attachments/files/113/566/606/034/369/077/original/a4d68d4ea151f24c.webp