OpenAI has slowed down the research and development of some cutting-edge models due to security issues. The new model Astra has reached the threshold of key network security capabilities.

#AIOpenAI As security concerns slow down the development of some cutting-edge models, the upcoming new model Astra has reached critical cybersecurity capability thresholds in its evaluation. OpenAI said that a variety of measures have been deployed internally to prevent model escape, and OpenAI is also strengthening security alignment. If the security environment, monitoring or alignment capabilities do not meet requirements in the future, model training may continue to be slowed down or even suspended. View the full text: https://ourl.co/126332