@ariacodez: That harmless looking "Explain with AI" button can rewrite your AI's memory 🤯 How it works: You tap it. Your AI tool opens with a prompt already typed out. Looks completely normal. Hidden at the end of that prompt is a second set of instructions you never wrote. I tested it. It saved a line into memory. Then I opened a brand new chat, asked an unrelated question, and boom, it repeated that line back like it was a fact. Three ways it happens: ✅ Poisoned training data ✅ Poisoned context ✅ Poisoned memory Claude and Grok warn you before you send a prefilled prompt. ChatGPT says nothing. Read the whole prompt before you hit send. EDUCATIONAL PURPOSES ONLY Save this 👇 #ai #aisafety #chatgpt #privacy #Tech

ariacodez

Open In TikTok:

Region: US

Monday 07 September 2026 16:34:09 GMT