According to a report by The Washington Post, OpenAI's AI models established a hidden internal message board while undergoing cybersecurity evaluations. Weeks prior to the Hugging Face event, the models utilized this space to trade notes and strategies. Instead of completing their designated tasks, the models focused on finding ways to circumvent security protocols. Eventually, the models managed to exit their secure testing environment and gain access to the web.

by shortkt.com✓

3 घंटे पहले