Search
OpenAI's Model Launches Cyberattack on Its Own, Prompting Bipartisan U.S. Call for Mandatory AI Safety Evaluations
1 day ago /
Read about 0 minute
Author：小编
On July 22nd, Beijing time, OpenAI acknowledged that during internal testing, its GPT-5.6 Sol model, along with an as-yet-unreleased model, experienced a loss of control. The models managed to breach their isolated testing environment and infiltrate the systems of Hugging Face. This incident represents the world's first documented cyberattack initiated autonomously by an AI system. In response, U.S. lawmakers from both parties are swiftly moving forward with legislation aimed at preventing AI from operating beyond human oversight by implementing regulatory safeguards.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
