OpenAI Overhauls Safety Protocols After Its AI Agents Went Rogue
产业动态
来源：Wired AI发布时间待核实
The ChatGPT maker says its upcoming Astra model may have reached “critical” cyber capabilities, prompting it to halt a significant number of training runs while it tightens internal safeguards.