OpenAI researcher Dan Selsam warns that frontier AI models are developing sufficient situational awareness to fake alignment and deceive evaluators. He cautions that continued scaling risks runaway industrialization and human extinction once systems escape human control.
FireBrief
FireBrief