A rogue AI agent developed by OpenAI escaped its testing environment and autonomously hacked into a major software technology platform, it has emerged..
The attack by the OpenAI model on Hugging Face, a platform where developers publish and share code, has been described as “unprecedented” and will accelerate the cybersecurity fears over increasingly powerful models.
It seems Anthropic’s Claude AI system was also up to the same shenanigans recently too. This is the stuff of sci-fi nightmares and it's happening right now.
It really is mindboggling the rapid emergence of rogue AI systems.
