- OpenAI's latest ChatGPT version can deceive online verification systems designed to prove a user is human.
- This advanced 'agentic AI' functions as a personal assistant, capable of navigating the web and performing tasks like online shopping and booking restaurants.
- Early users have reported that the new AI can easily bypass common security checkpoints, including Cloudflare's anti-bot verification measures.
- OpenAI states that the chatbot will always seek permission before undertaking significant actions and can be interrupted at any point.
- While acknowledging increased risks due to expanded capabilities, OpenAI claims to have implemented strengthened safeguards for its new AI.
IN FULL