New ChatGPT Atlas Browser Exploit Lets Attackers Plant Persistent Hidden Commands
... artificial intelligence ( AI )-powered assistant's memory and run ... "What makes this exploit uniquely dangerous is that it targets the AI's ...
Read the original at The Hacker News ↗
- Topics
- AI Safety & Existential Risk
- Mentions
- ChatGPTOpenAI
- Tracked via
- “dangerous AI”