OpenAI Hits the Brakes: Pausing Frontier RL Training Over “Astra” Cyber-Critical Risk
中文摘要
OpenAI 因发现重大网络安全风险，已暂停其名为“Astra”的前沿强化学习模型训练。
English Summary
OpenAI paused training its "Astra" frontier reinforcement learning model after discovering significant cyber-critical risks during the development process.
Original Excerpt
If you build anything on top of AI models, this is the story to actually understand — not just skim past. Continue reading on Medium »